United States
Environmental Protection
Agency
Office of Research and
Development
Washington, DC 20460
March 1993
Conference on the Risk
Assessment Paradigm
After Ten Years:

Policy and Practice Then,
Now, and in the Future
April 5-8, 1993
Hope Hotel and Conference Center
Dayton, Ohio

-------

-------
                                                   EPA/630/R-93/039
                                                       March 1993
   CONFERENCE ON THE RISK ASSESSMENT PARADIGM AFTER TEN YEARS;
        POLICY AND PRACTICE THEN, NOW, AND IN THE  FUTURE
                      BIOGRAPHIES AND PAPERS
                             SPONSORS


    Toxicology Division, Occupational  and Environmental Health
Directorate, Armstrong Laboratory

     Naval Medical Research Institute  Detachment (Toxicology)

       Army Biomedical Research and Development Laboratory

 The U.S.  Environmentcil Protection Agency, Environmental Criteria
                       and Assessment Office
                        In Cooperation with

      The National Research Council Committee on  Toxicology
                                                    Printed on Recycled Paper

-------

-------
            BROTHER, TIMES ARE TOUGH; CAN YOU PARADIGM?

                             Dr. Donald G. Barnes
                                Staff Director
                     Science Advisory Board of the USEPA
OUTLINE OF THE TALK

 I.  Introduction

    A.  The Paradigm

       1.  Risk Assessment

          a. "Is this stuff toxic?" - Hazard identification
          b. "How toxic is this stuff?" - Dose/Response assessment
          c. "Who is exposed to this stuff, to how much, how often, and for
            how long?" - Exposure assessment
          d. "So what?" - Risk characterization

       2.  Risk Management:  "So what are you going to do about it?

    B.  The Talk

       1.  Impact and strengths of the paradigm
       2.  Strains and weaknesses of the paradigm
       3.  Where do we go from here?
    The Impact and Strengths of the MAS Approach

    A.  Succeeded in nailing a lot of - but not all of - Jello to the wall

       1. Common, somewhat demystified lexicon

         a. For anointed practitioners
         b. For laypeople

       2. Conceptual separation of church (RA) and State (RM)

       3. Broad application to different situations

     •  4. Distinguishes between areas of fact and "faith," shining a research light
         into the darkness

                                     -1-

-------
    B. Evidence of its Utility

       1. Wide adoption of the concepts; e.g.,

         a. EPA guidelines

         b. IRIS

         c. Other agencies; e.g., HHS, where an official once denigrated
            risk assessment (not totally without basis) as being "as accurate
            as a five-year weather forecast"

         d. Innumerable articles, books and conferences

       2. Structured, disciplined decision-making that can distinguish between
         RA and RM, relieving the burden on the scientist while increasing the
         burden on the risk manager


III.  Strains and Weaknesses of the MAS Approach

    A. Total separation of RA and RM is not possible - or even desirable in many
       instances

    B. As used, the paradigm favors a reductionist - single chemical (stressor) -
       approach; i.e., prejudices us against addressing mixtures in a holistic manner

    C. Current approach does not address adversity of effect

    D. As used, there is ambiguity about whether hazard identification should relate
       to "exposure under any conditions" or "exposure likely to be encountered in
       the  environment"

    E. Ecological paradigm  is purported to be different
 IV. Where "do we go from here?

    A. Dynamic tension in the issues in III probably precludes their final resolution
       in the short term; however, they should be addressed at the technical and
      • policy levels (Review each of the points and give recommendations)

    B. The long-awaited NAS "update"

  V. Conclusions and Recommendations

                                       -2-

-------
DISCUSSION OF SELECTED ISSUES
III.  Strains and Weaknesses of the NAS Approach

    A. Total separation of RA and RM is not possible - or even desirable in many
       instances

       Some degree of interaction between RA and RM is essential if the RA answers are
going to address - let alone satisfy - the RM questions.  In some instances, for example,
a qualitative answer will suffice; e.g., "Could the pollutant run off into the stream and
.bioaccumulate  in  aquatic organisms?"  In other instances, a more detailed answer  is
neecjecj; e.g., "What are the remedial options that would prevent runoff of the pollutant to
such an extent that bioaccumulation would be maintained at levels below a 10~5 risk level
to the sportsfishing  population?"

       In fact, the NAS considered the option of separating the RA operations totally from
the RM operations; e.g., establish a separate agency to conduct such analyses. To their
credit, the NAS panel rejected this option  as  being infeasible, while spotlighting the
.importance of separating RA and RM functions within'a single agency.

       In finding  the  proper balance, there  will - and should - be a continual tension
between the  need for good communication between the customer (RM) and supplier
(RA).                                  ,


     B. As used, the paradigm favors a reductionist  - single chemical (stressor) -
       approach; i.e., prejudices us against addressing mixtures in a holistic manner

       After an initial confusing and often tense several years after the NAS (1983) report,
risk assessors and risk managers  have mgje  or less sorted their respective roles and
responsibilities under the paradigm. This confusion, tension  and resolution has proved
very useful for environmental protection.  It has, for example, allowed the U.S.  EPA to
form a Risk Assessment Forum, a Reference Dose (RfD)/Reference Concentration  (RfC)
Work Group and a Carcinogen Risk Assessment Verification Endeavor (CRAVE) Work
 Group. These three groups focus on risRassessment guidelines, methods, and chemical-
 specific evaluations on an intra-agency basis. Evaluations and recommendations of these
 groups are then  used by risk, managers of different program offices in  their decision
 process.
                                       -\
                                       -3-

-------
       Risk assessment and risk management have been practiced most often on a
chemical-by-chemical basis.   One might say that the paradigm even favors  such an
analysis  since  initial  steps  in the paradigm (i.e.,  hazard Identification,  exposure
assessment) operate  most  readily on a single  chemical  basis.  Unfortunately, our
environment seldom offers us or our ecosystems exposures to single chemicals.  Future
shins in the paradigm must stress this  reality, and guide both assessors and managers
into credible approaches to this mixtures problem.


     C. Current approach does not address adversity of effect    .

       A perennial chestnut, the issue of adversity of effect - e.g.,  "Which is  worse:
cancer or development effects?" - has withstood all attempts at definitive resolution by
the Agency, the MAS, and anyone else for that matter.  At bottom, the question of the
relative  concern of leukemia vs. missing limbs or reproductive effects vs. stratospheric
ozone depletion appears to be a "value judgment" to be made somewhere else other than
\n the RA arena While such a statement may be true, it reflects a limitation in the ability
of RA to answer the RM question: "What are you going to do about these two risks?"

        There is increasing recognition that resources to address environmental problems
are limited  Consequently, trade-offs have to be made in many cases, including those in
which a variety of MAS risks estimated via the MAS paradigm  are generally equaL  In
order to reach decisions in such cases, therefore, something beyond the MAS paradigm
is needed.
     D  As used, there is ambiguity about whether hazard identification should relate
        to "exposure under any conditions" or "exposure likely to be encountered in
        the environment"

        As  most often used under the paradigm,  hazard identification focuses  on
 laboratory  investigation of  toxic effect, whether looking  at experimental animals  as
 surrogates for humans, or at single species as surrogates for a community or ecosystem.
 Epidemiological or field investigation also .serves a very useful role in identifying hazards,
 especially when effects can be clearly ascribed to the stressor.  Scientists restrict such
 laboratory, epidemiological or field investigation to limited exposures. By necessity, these
 exposures are then generalized to environmental  situations.

        One purpose for such restriction is to obtain unequivocal results.  Since hazard
 identification  often starts a more comprehensive risk assessment resulting in a risk
 management  decision, .unequivocal  results are highly valued.  Unfortunately, such
 unequivocal  hazard identification  sometimes (often?) has  limited relevance  to  an
 environmental exposure,  since environmental exposures are often to mixtures or forms
 of the  chemical that were not tested.

-------
       The current paradigm supports this dilemma.  Improvements in the paradigm
should discuss what, if anything, can be done about this.
    E. Ecological paradigm is purported to be different

       Recently, the Agency  issued  an  "Ecological Risk Assessment  Framework"
document containing a paradigm for eco-RA that is somewhat different  from that
proposed by the NAS a decade ago. The reasons for these differences are several; e.g.,

    •  We know more about RA than we did 10 years ago

    •  The ecological problems are qualitatively different from the health problems -
       that were the focus of the original NAS concerns

    •  The ecological community wanted to leave their distinctive mark along the RA
       trail    •;•       .

       In any event, there should be a single RA paradigm that is sufficiently broad to
encompass both health and ecological  concerns.  Currently, there is the danger that
eco-RAers and health-RAers will evolve in different - and possibly opposing - directions.
     The views expressed in this paper are those of the author and do not necessarily
 represent thfee of EPA or the Science Advisory Board.
                                      -5-

-------
BIOSKETCH:  LAWRENCE W. BARNTHOUSE
Dr   Barnthouse  has  been  a  staff  member  in the  Environmental
Sciences Division (BSD)  at Oak Ridge National Laboratory (ORNL) for
16 years,  and has been leader  of ESD's  Environmental Risk Group
since  1983.   He received his A.B.  Degree  in Biology from Kenyon
College  in 1968 and his Ph.D.  in Biology  from the University of
Chicago  in 1976.  He has led  or participated  in a wide variety of
ecological research and assessment projects involving common themes
of  (1)  extrapolating from  laboratory to field,  (2)  modeling of
population and ecosystem responses to environmental stress, and (3)
relating  scientific  information  to  regulatory  needs.    Major
problems addressed have included impacts  of power plant cooling
systems  on estuarine fish populations  and communities; ecological
risks  of  synthetic  fossil  fuel technologies; and evaluation of
remedial action priorities  at contaminated  sites.

Dr.  Barnthouse also serves as Deputy Director of the  ORNL Center
for  Risk  Management.   He  is a member  of  the National Research
Council's Committee on Risk Assessment Methodologies,  where he
chairs the Ecological  Risk Topic Group.  He  has written numerous
publications on ecological  risk assessment and was  named Marietta
Energy Systems Author of the Year  for 1991.   In November 1992 he
became Hazard  Assessment Editor of Environmental  Toxicology and
Chemistry.

-------
             Issues in Ecological Risk Assessment:  the  CRAM Perspective
                                  Lawrence W. Barnthouse
                               Environmental Sciences Division
                               Oak Ridge National Laboratory
                              Box 2008, Oak Ridge, Tennessee1
1 Managed by Martin Marietta Energy Systems, Incorporated under Contract DE-AC05-84OR214QO with
the U.S. Department of Energy

-------
                                           Abstract
 In 1989, a  Committee on Risk Assessment Methodology was convened by the National Research
 Council'to identify and investigate important scientific issues in risk assessment.  One of the first issues
 considered by the committee was the development of a conceptual framework for ecological risk
 assessment, defined as the characterization of (he adverse ecological effects of environmental exposures to
 hazards imposed by human activities..  "Adverse ecological effects" include all biological and
 nonbiological environmental changes that society perceives as undesirable. The committee's opinion was
 that a general framework is needed to define the relationship of ecological risk assessment to
 environmental management and to facilitate the development of uniform technical guidelines.  The
 framework for human health risk assessment proposed by the NRC in 1983 was adopted as a starting
 point for discussion.

 CRAM concluded that, although ecological risk assessment and human health risk assessment differ
 substantially in terms of scintific disciplines and technical problems, the underlying decision process is the
 same for both.  CRAM therefore recommended that the  1983 risk assessment framework be modified to
' accomodate both human health and ecological risk assessment. CRAM defined an integrated
  health/ecological risk assessment framework consisting  of the four components Hazard Identification,
  Exposure Assessment, Exposure-response Assessment,  and Risk Characterization. CRAM further
  provided recomendations on the scope of issues to be addressed in ecological risk assessment, critical
  research needs, and mechanisms for providing more detailed guidance on the scientific content of
  ecological risk assessments.

-------
                                         Introduction

       In 1983 the National Research Council's Committee on the Institutional Means for Assessment of
Risks to Public Health published a landmark report on human health risk assessment. The report, Risk
Assessment in the Federal Government:  Managing the Process [I], proposed a conceptual framework
for risk assessment that incorporates research, risk assessment, and risk management.  Risk assessment
was defined as "...the characterization of the potential adverse health effects of human exposures to
environmental hazards." The report proposed a conceputal scheme for risk assessment consisting of four
components: hazard identification, dose-response assessment, exposure assessment, and risk
characterization. The report did not, however, include in-depth discussion of scientific issues' in health
risk assessment.  The 1983 committee's objectives were limited to addressing institutional and procedural
issues: whether the analytic process of risk assessment should be cleanly separated from the regulatory
process of risk management, whether single organization could be designated to perform risk assessments
for all regulatory agencies, and whether uniform risk assessment guidelines could be developed for use by
all regulatory agencies. Detailed development of technical guidelines was left to the agencies themselves.
       In 1989, a new Committee on Risk Assessment Methodology  (CRAM) was convened within the
 Board on Environmental Studies and Toxicology of the National Research Council's Commission on Life
 Sciences to identify and investigate important scientific issues in risk assessment. The committee was
 asked to consider changes in the scientific foundation of risk assessment that have occurred since the 1983
 report and to consider applications of risk assessment to non-cancer endpoints.  The first three issues
 considered by CRAM were (1) the use of the Maximum Tolerated Dose in animal bioassays, (2) the use
 of the two-stage model of carcinogenisis, and (3) the development of a concetpual framework for
 ecological risk assessment The committee has recently issued a report on these three issues [2]. In
 addition to describing an integrated framework for human health and ecological risk assessment, CRAM's
 report discusses the scope of applicability of ecological risk assessment and identifies major categories of
 scientific uncertainty for which additional research is needed.  The purpose of this paper is to briefly

-------
describe the framework recommended by the Committee and compare it to EPA's recently-published
Framework for Ecological Risk Assessment [3, 4].

       F.rn1n?ical risk assessment was defined by CRAM as the characterization of the adverse
ecological effects of environmental exposures to hazards imposed  by human activities.
"Adverse ecological effects" include all biological and nonbiological environmental changes that society
perceives as undesirable. "Hazards" include both unintentional hazards such as pollution and soil erosion
and deliberate management activities such as forestry and fishing that are often hazardous either to the
managed resource itself or to other components of the environment. The committee's opinion was that a
general framework analogous to the 1983 human health risk assessment framework is needed to define the
relationship of ecological,risk assessment to environmental management and to  facilitate the development
of uniform technical guidelines. A framework for ecological risk assessment could, for example, be used
to:

                     Evaluate the consistency and adequacy of individual assessments,
                     Compare assessments for related environmental problems,
                     Identify explicitly the connections between risk assessment and risk management,
                     and
                     Identify environmental research topics and data needs common to many ecological
                     risk assessment problems.

  Like the health risk assessment framework, an ecological risk assessment framework would define the
  boundaries between risk assessment and risk management and identify general categories of scientific
  information relevant to risk assessment, but would not provided specific technical guidance.

-------
       The committee chose to investigate the feasibility issue by conducting a workshop in which six
case studies representing different types of current assessments would be examined with respect to their
          t
consistency with a common framework. The six case studies included:

              •       Assessing the effects of tributyltin on Chesapeake Bay shellfish populations,
              •       Testing agricultural chemicals for ecological effects
              •       Predicting the fate and effects of polychlorinated biphenyls,
              •       Assessing responses of populations to habitat change,
              •       Regulating species introductions, and
              •       Harvesting the Georges Bank multispecies fishery.
        A Workshop on Ecological Risk Assessment was held on February 26-March 1,1991, at Air lie
House, Warrenton, Virginia. In addition to presentation and discussion of the case study papers, the
workshop included breakout sessions to discuss conceptual and technical aspects of ecological risk
assessment. A summary of the workshop presentations and dicsussion is included as an appendix to the
CRAM report [2]; three of the case study papers have been independently published [5, 6, 7]. A
general consensus emerged at the workshop that an ecological version of the 1983 framework is desirable
and feasible, but no specific endorsement of a particular framework was sought or obtained.  On
reviewing the written materials produced at the workshop, the committee concluded that the 1983 human
health framework could be expanded to accomodate both human health and ecological risk assessment.
For general applicability to ecological assessments, the 1983 scheme requires augmentation to address
some of the interfaces between science and management, primarily because of the need to focus oh
appropriate questions relevant to applicable environmental law and policy under different circumstances.
Specifically, the scheme needs modification to address  (1) the influence of legal and regulatory
considerations on the initial stages of ecological risk assessment and (2) the importance of characterizing
ecological risks in terms that are intelligible to risk managers.  The committee's opinion is that these

-------
augmentations are as
assessment.
important for human health risk assessment as they are for ecological risk
                                 The Integrated Framework
       CRAM concluded that integration of ecological risks into the 1983 risk assessment framework is
 preferable to developing a de nova ecological risk assessment framework. Like health risk assessment,
 ecological risk assessment must be defined in broad terms if it is to be applicable to the full array of
 environmental problems that regulatory and resource management agencies must address. Moreover, any   -
 framework chosen for ecological risk assessment must be simple, flexible; and general, so that it will be
 understood by both scientists and the risk managers with whom scientists must communicate.  The 1983
 framework, by any measure, has been extraordinarily successful in communicating the broad features of
 health risk assessment throughout the scientific and regualtory communities.  Although ecological risk
 assessment and human health risk assessment differ substantially in terms of scintific disciplines and
 technical problems, CRAM concluded that the underlying decision process is the same for both. The
 function of risk assessment is to link science to decisionmaking, and that basic function is essentially the
 same whether it is risks to man or risks to the environment that are being considered.

         The  1983 report defined ^^ identification as "...the process of determining whether exposure
  to an agent can cause an increase in the incidence of a health condition," including "...characterizing the
  nature and strength of the  evidence of causation." Doy-re.sron.se assessrneqt was defined as "...the
  process of characterizing the relation between the dose of an agent administered or received and the
  incidence of an adverse health effect as a function of human exposure to the agent," accounting for
  exposure intensity, age, sex, lifestyle,and other variables affecting human health resposes to hazardous
  agents. *vr-— assessment was defined as "...the process of measuring or estimating the intensity,
  frequency, and duration of human exposures to an agent currently present in the environment or of
  estimating hypothetical exposures that might arise from the release of new chemicals into the

-------
environment" Risk characteriTatinq was defined as "...the process of estimating the incidence of a health
effect under the various conditions of human exposure described in exposure assessment It is performed
by combining the exposure and dose-response assessments. The summary of effects of the uncertainties
in the preceding steps are described in this step."

       On evaluating the consistency of the six case studies presented at the workshop with the 1983
framework, CRAM concluded that most of the case studies fit reasonably well. The most obvious
common deficiency related to risk characterization. Only one of the case studies, the Georges Bank study,
included any quantification of risks in terms that could be used for risk-benefit calculations, valuation
studies, or other quantitative comparisons applicable to decision-making. Even in this case, the value of
the assessment to decisionmaking is uncertain. During plenary discussion,  the study author emphasized
that communication between scientists and managers is still inadequate and that fisheries management
actions are often only marginally influenced by quantitative assessments. Approaches to hazard
identification exemplified in the case studies were, on  the other hand, substantially more diverse and in
some cases more sophisticated than  envisioned in the 1983 framework. Ecological hazard identifications
often include identifications of specific species or ecosystems of interest, delineation of study areas, and
determination of types of laboratory or field data on which an assessment will be based. These decisions
reflect both scientific considerations (which systems are vulnerable? what kinds of effects are possible?)
and management considerations (which species or ecosystems are to be protected? must costs be weighed
against benefits? is the objective to protect the resource or to optimize exploitation of the resource?). The
workshop consensus was that definitions of hazard identification and risk characterization proposed in the
1983 report are inadequate for the purposes of ecological risk assessment
       CRAM agreed with the consensus at the workshop that the 1983 framework is inadequate, as
written for application to ecological problems, because (1) it does not account for legal mandates and other
policy, considerations that influence the initial stages and focus of ecological risk assessments, and (2) it
pays insufficient attention to the critical problem of effective commun

-------
public. The opinion of the committee, however, is that these deficiencies are not unique to ecological risk
assessment. Differences in the functions of different regulatory agencies clearly influence the types of data
and inference guidelines used in health risk assessments, and effective risk communication is as important
(and often as inadequately performed) in health as in ecological risk assessment.

       Ha.^ identification was redefined by CRAM to be the determination of whether a particular
hazardous agent is associated with health or ecological effects of sufficient importance to warrant further
scientific study or immediate management action. Fvpnsnr^sponse wwv* was defined as the
determination of the relation between the magnitude of exposure and the probability of occurrence of the
effects in question. Replacement of the term "dose" with a more general term is required, because "dose-
 has a distinctly medical connotation and cannot be effectively applied to nonchemical stresses, such as
 habitat change or harvesting.  The "responses" addressed in ecological risk assessments include both
 direct effects of exposure and the much broader indirect effects, such as secondary poisoning of raptors
 due to accumulation of pesticide residues in their prey and effects of harvesting on fish-community
 structure.  Exp^m^SSSsmsiit  was defined byCRAM as the determination of the extentof exposure to
 the hazardous agent in question before or after application of regulatory controls. In the committee's view,
 the term "exposure" can legitimately be applied to nonchemical stresses, including both physical stresses
 (such as habitat change and UV radiation) and biological stresses (such as species introductions).
 Alternative terms, e.g., "stress" or "stressor"  were deemed unsuitable because of conflicts with medical
 uses of the same or similar terms. Risk characterisation.  was defined as the description of the nature and
  often the magnitude of risk, including attendant uncertainty, expressed in terms that are
  comprehensible to decisionmakers and the public.

         The revised framework is summarized in Figure 1.  In addition to the four basic components,
  Figure 1  depicts two key aspects of risk assessment  As noted above, it is essential to recognize the
  influence of policy considerations on hazard identification. CRAM also wanted to emphasize the need to
  create a connection between the results of today's risk assessments and the science base for future risk

-------
assessments. The risk assessment process should not end when a regulatory decision is made. Followup
in the form of monitoring (where measurable effects have been predicted), validation studies, and basic
research are needed to improve the data and models available to technical risk assessors whenever the same
or a similar problem is encountered in the future.

            Comparison to EPA's  Framework for  Ecological Risk  Assessment

       EPA's recently-published "Framework for Ecological Risk Assessment" [3, 4] is quite similar to
CRAM's integrated framework, and the similarity is not accidental. EPA consciously modeled its
framework on the 1983 NRC health risk assessment framework. Moreover, several of the authors of
EPA's framework document participated in the CRAM ecological risk assessment workshop and a CRAM
member served on a review panel that evaluated EPA's framework [8]. CRAM's "hazard identification"
is replaced by problem formulation in EPA's version. CRAM's "exposure assessment" and "exposure-
response assessment"  are subsumed by EPA in a step called Analysis, which is in turn subdivided into
Characterization of Exposure and Characterization of Effects  Definitions of the components are more
specifically ecological and somewhat more explicit than are  the definitions in the CRAM framework.
Problem formulation,  for example, is described as a "systematic planning step" that includes discussions
with risk managers, preliminary description of the potential ecological effects of the stressor, identification
of the specific effects (termed " assessment'endpoints") to be addressed in the assessment, and
development of a conceptual model to guide the assessment
       The relationship between assessment and management in the EPA framework reflects EPA's
specifically regulatory mission and might be approached differently by another agency or a private-sector
organization involved in ecological risk assessment  Policy input is provided by a risk manager who
discusses the assessment with the technical staff during the problem formulation phase. When the
assessment is complete, the results are discussed with the risk manager, who then is responsible for
making a decision and communicating the results to the public at large.  In more general kinds of

-------
assessments, such as environmental impact assessments performed to satisfy the National Environmental
Policy Act, the planning phase (termed "scoping" in NEPA regulations) includes substantial public
involvement. In others, such as assessments performed during development of environmentally safe
products, frequent iterative interactions between design engineers, marketing staff, and risk assessors
might be expected.

                          The future of ecological risk assessment

 Neither the CRAM framework nor the EPA framework were intended to provide an explicit recipe for the
 scientific content of ecological risk assessment. The EPA expects the process of technical guidance
 development to implement its framework to take several years [4]. The CRAM report recommends that
 expert committees be convened to discuss the major scientific issues in ecological risk assessment. The
 report identifies four major areas in which scientific consensus is lacking: extrapolation across scales of
 time, space, and ecological organization; quantification of uncertainty; validation of predictive tools;  and
 economic valuation of ecological resources. The principal objective of both frameworks is to provide a
 common conceptual foundation that can enhance the consistency and credibility of ecological risk
 assessments.

        CRAM made five specific recommendations concerning the future development and use of
 ecological risk assessment.:

               •Risk assessors, risk managers, and regulatory agencies should adopt a uniform
                    framework for ecological risk assessment The extension of the 1983 NRC human
                    health risk assessment framework described in the CRAM report and depicted in
                   . Figure 1 is general enough to apply to most assessment problems and emphasizes the
                    common elements of health risk and ecological risk assessment.
                                                 10

-------
              •State and federal agencies should expand the issue of risk assessment in strategic planning
                  and priority-setting as a means of focusing their resources on critical environmental
                  problems and uncertainties.

              •Agencies should support the development of improved methods of risk characterization
                  and consistent, guidelines for applying them. Specific areas where current approaches
                  are inadequate include extrapolation of population and ecosystem effects, expression
                  of risks in terms that are useful for decision-making and understood by the public at
                  large, and evaluation and communication of both quantitative and qualitative
                  uncertainties.

              •To improve the science base for future risk assessments,  agencies should institute
                  systematic followup of risk assessments with research and monitoring to determine the
                  accuracy of predictions and resolve remaining uncertainties.

              •EPA and other agencies should support systematic research programs to improve the
                  credibility and utility of ecological risk assessments, and should draw on scientific
                  expertise available outside the agencies themselves to develop technical guidance on
                  the scientific content of ecological risk assessments.   „

       The intent of CRAM's recommendations is to facilitate understanding of ecological assessment
principles by nontechnical decisionmakers and the public at large and to ensure consistent improvement in
the science supporting ecological risk assessment. The past few years have seen a major increase in public
interest in the environment The adoption of "sustainable development" as general environmental goal
implies that economic development strategies should strive to simultaneously maximize both human
welfare and environmental quality.   An integrated framework for risk assessment of the kind
recommended by CRAM can facilitate achievement of this goal.
                                                 11

-------
                                    Acknowledgements

The author gratefully acknowledges the contributions of (1) all members of the National Research Council
Committee on Risk Assessment Methodology, particularly Chairman Bernard Goldstein and Ecological
Risk Assessment Topic Group members Alan Maki and Warner North, and (2) the staff of the Board on
Environmental Studies and Toxicology, particularly Linda Leonard, James Reisa, Richard Thomas, Mary
Paxton, Kathleen Stratton, and Gail Charnley.  Oak Ridge National Laboratory is managed by Martin
Marietta Energy Systems, Inc., under contract DE-AC05-84OR 21400 with the U.S. Department of
Energy.
                                               12

-------
References
National Research Council (NRC). Risk Assessment in the Federal Government: Managing the Process.
     ,  National Academy Press, Washington, D.C. (1983)

National Research Council (NRC).  Issues in Risk Assessment.  National Academy Press, Washington,
       D.C.(1993)

Environmental Protection Agency (EPA), Framework for ecological risk assessment. Rep. No.
       EPA/630/R-02/001, U.S. Environmental Protection Agency, Washington, D.C. (1992).

Norton, S.B.,  D. J. Rodier, J.H. Gentile, W.H. van der Schalie, W.P. Wood, and M.W. Slimak. A
       framework for ecological risk assessment at the EPA.  Environmental Toxicology and Chemistry
       11:1663-1672 (1992).

Fogarty, M. J., A:  A. Rosenberg, and M. P. Sissenwine. Fisheries risk assessment: sources of
       uncertainty. Environ. Sci. & Technol 26:440-447 (1992).

Huggett, R. J., M.  A. Unger, P. F. Seligman, and A. O. Valkirs.  The marine biocide tributyltin.
       Environ. Sci. & Technol. 26:232-237 (1992).
                                *            , '
Kendall, R. J.  Farming with agrochemicals: the response of wildlife. Environ. Sci. & Technol. 26:239-
       245 (1992).
                                             13

-------
Environmental Protection Agency (EPA). Peer review workshop report on a framework for ecological



      risk assessment.  Rep. No. EPA/625/3-91/022, U.S. Environmental Protection Agency,




      Washington, D.C. (1992)
                                               14

-------
BIOGRAPHICAL SKETCH: GLEN J. BARRETT, C.I.H.
Mr. Barrett received his undergraduate degree in Chemistry from Marquette University
in  1968.  He received his Master's in  Inorganic  Chemistry, also from Marquette
University, in 1970.

Mr. Barrett is currently a Certified Industrial Hygienist and Health and Safety Officer
with The Earth Technology Corporation.  He has more than 15 years of experience in
all aspects of industrial  hygiene services, including  health  and safety  program
development and implementation, preparation of hazardous substance fact sheets for
state hazard communication programs, preparation of human health risk assessments,
health and safety training, control technology assessment, and industrial hygiene
program management. He is responsible for development and implementation of Earth
Technology's health and safety programs and management of  risk  assessment
projects. He has managed air monitoring for major asbestos abatement projects, and
has  presented asbestos worker, supervisor,  inspector, and management planner
Asbestos Hazard Emergency Response Act (AHERA) training.

Mr."Barrett has developed health and safety plans for the Environmental Protection
Agency, Region 3 and for the U. S.  Air Force Center for Environmental Excellence
under the Air Force Installation Restoration Program, and has managed human health
risk assessments for various locations.

-------
                      Hazard
                    Identification
                                   Exposure:
                                   Response
                                   Assessment
 Exposure
Assessment
                i
                  Risk
             Characterization
Science
 research
validation
monitoring
Figure 1. The CRAM integrated human health/ecological risk assessment framework

-------
           A Computer-Aided Approach to Quantification of Human Intake
                              for Risk Assessments

                         Glen J. Barrett, M.S., C.I.H.,
               Leonard M.  Pried,  M.S.,  and Jennifer A. Smith,  M.A.
                        The Earth Technology Corporation
                              Alexandria, Virginia
                    A Computer Approach for Risk Assessments
                                Glen J. Barrett
                        The Earth Technology Corporation
                           1420 King Street, Suite 600
                           Alexandria,  Virginia 22314
023«.rcv

-------
ABSTRACT

      A Risk Assessment  was conducted by The Earth Technology Corporation to
.valuate risks to human  health  resulting  from soil contamination found at an
evaluate *«*                            the use of several dBase programs that
rapid review of a large quantity of data.
      A total of 858 soil samples were collected and analyzed  for 25 inorganic

          r: rrr                      r r r -;=
 Restoration Program Information Management System (IRPIMS)  format.

      A methodical approach was used to select chemicals of potential concern for
 risk calculations.  Unusable data were deleted from the IRPIMS databases «th the-
 aid of several  dBase  programs.   A  dBase  program,  SITESTAT.PRG,  was used to
 calculate statistical data.  This program was executed separately on «*-dual
 site databases  for all depths sampled and for specific depths -of -concern   Data
 generated by SITESTAT.PRG included maximum concentration -detected  ar.thmet.c
 fa:! ^JL**-.  *~ »-«  of  valid detections  and the_ total  number of
 samples analyzed.   Analytical data were evaluated wxth a variety of criteria
 including frequency and magnitude of detection and consideration of the depth of
 concern.  A Sase program, ABOVEMAX.PRG,  was  developed to greatly ^^
 comparisons of  site  inorganic  data to background  site data.   The col ect ve
 result of these analyses is a table in which a  summary is provided to justify
 retlntion or deletion of a chemical of concern for  further evaluation for  each
 site .

        Human receptors who could be impacted by  migration of site contaminants or
 by direct contact with site soil were identified.  Soil contact exposure .pathways
 were  identified for  these  receptors.   Two  dBase programs,  AIREXPS.PRG and
 SOIBEXP.PRG, were .developed   by The  Earth Technology Corporation to estimate
 average  exposure  and reasonable maximum exposure  for identified axr and  so.l
 exposure pathways.

 KEY WORDS:  Risk,  Intake, Exposure, dBase
  023d rev

-------
INTRODUCTION

      A  baseline  risk assessment  was  conducted  as part  of a  remedial soil
investigation to  evaluate  baseline risks to human  health  resulting from soil
contamination found at an industrial facility,  A total of 858 soil samples were
collected from 10 potential hazardous  waste  sites  and analyzed for 25 inorganic
and 159 organic chemicals.   A risk  assessment, was  not conducted for groundwater
•contamination because:

      •     The  regional aquifer  which supplies drinking water for  offsite
            residents  has  been identified  as  contaminated,  and  an operating
            remediation system is in effect

      •     The  remediation  system  has impeded  migration of  contaminated
            groundwater off the industrial facility property

      •     Current and  future  offsite residents will not  be impacted by the
            contaminated plume contained on the facility property.

      •     Uncontaminated drinking water is supplied to facility workers by a
            municipal water system.

      Several references provided  guidance  for the baseline risk assessment.
These references included:

      •     Risk Assessment  Guidance  for Superfund,  Volume  I:   Human Health
            Evaluation  Manual  (Part  A),  Interim  Final,   U.S.  Environmental
            Protection  Agency  (USEPA),  Office  of  Emergency  and  Remedial
            Response, Washington,  D.C., December  1989
      •     Risk Assessment  Guidance for Superfund,  Volume I:   Human Health
            Evaluation  Manual,  Supplemental  Guidance,  "Standard  Superfund
            Default Exposure Factors", Interim Final, USEPA,  Office of Emergency
            and Remedial Response, Washington,  D.C., March 1991.

      Soil analytical data derived from the  sampling effort were available in a
set of  related databases in the  Installation  Restoration .Program Information
Management System  (IRPIMS)  format.   These  databases  were constructed as .DBF
files and were managed using dBase III Plus version 1.0.  DBase was chosen for
its applicability and cost effectiveness.

      This paper describes  the use  of several  dBase programs  to calculate

-------
statistics used for the selection of chemicals of potential concern, to calculate
the average and 95* upper confidence limit (UCL)  concentrations for chemicals of
potential concern at  identified human receptors, and to estimate  human daily
intake values for identified pathways,   Stacistical results from these programs
greatly facilitated rapid review of a large quantity of data.

     Although a full,  baseline risk assessment was conducted,  only the following
risk assessment procedures will be discussed:

      •     Selection of Chemicals of Potential Concern
      •     Identification of Human Receptors
      •     Identification of Exposure Pathways
      •     Estimation of Concentrations  of chemicals  of potential concern at
            human receptors
      •     Estimation of human intake values.

OVERVIEW OF DATABASE MANAGEMENT

      The IRPIMS format required the use of several databases; BCHRES.DBF (batch
result database)  was  the main database used. It contained the analytical  soil
results for each analysis performed on all samples. Specifically, each BCHRES.DBF
record contained the analytical results for a single chemical by sample; a trip,
equipment,  ambient  air,  or  laboratory blank;  or for  a  surrogate  spike.

      BCHRES.DBF  was used to  construct an individual site database for each site
evaluated.  BCHRES.DBF is composed of many fields;  only a subset of these fields
were  used for the risk assessment. Table I displays  the  BCHRES.DBF fields used
for the  risk assessment.

SELECTION OF CHEMICALS OF POTENTIAL CONCERN

       The following criteria were used to initially screen organic and inorganic
 soil  data:

       1     Data  Validation.    Ten  percent  of the  data  were  validated  in
 accordance with USEPA guidelines (1) (2) .  All qualified data were reviewed. Table
 II presents validation qualifiers used for soil data.

      A dBase program inserted  'R' or  'DR' qualifiers in the RAQUALIFY field in
 BCHRES.DBF."Any records qualified with a 'R' or 'UR' were considered unusable and
 02JS.tr*

-------
                                   Table X        .
                BCHRES.DBF Fields Used for the Risk Assessment
      LOCXREF    -   A unique  identifier  for the  location of the sample;
                    typically,  the borehole ID
      ANMCODE    -   The analytical method code
      EXMCODE    -   The extraction method code
      SBD        -   The sample  beginning depth
      SED        -   The sample  ending  depth
      PARLABEL    -   Chemical  label
      PARVQ       -   Indicator of  sample  detection
      PARVALDLUN -   Dry weight  soil  concentration (mg/kg)
      LABDL       -   Laboratory  detection limit
      SACODE     -   Type of sample  (blank,  field replicate)
      SITEID     -   Number designating the site  from which the sample
                    was obtained
      RAQUALIFY  -   Validation  qualifier (U,UJ,J,UR,R)
      HTFLAG     -.  Indicates the sample exceeded a holding time
      NDBLANKC    -   Indicates the sample did not exceed 10 or 5 times,
                    depending on  chemical,  the maximum detection of  any
     	blank associated with the sample batch	
023«.iCT

-------
                                  Table  II
                       Soil Data Validation Qualifiers
    -J	IThe associated numerical value  is  an estimated quantity

    R     -    The data are unusable  (compound may or may not be present) ;
              Resampling and reanalysis  is  necessary for verification

    CTJ-   -    The material was  analyzed,  but  the analyte was not detected;
              The sample quantitation limit is an estimated quantity

    UR   -    The material was  analyzed,  but  the analyte was not detected;
              The data are unusable  and  is  rejected

    tr    -    The material was  analyzed,  but the analyte was not detected;
              The associated value is the sample quantitation limit	
OZS&tcv

-------
were, not added to any individual  site database. A data qualified with 'J'
or 'UJ' were considered valid and were retained in site databases.
'U'
      2.       Quality Control  Analysis,    A  dBase program  was  developed  to
evaluate the analytical results for all quality control blanks in accordance with
USEPA Contract Laboratory Program  (CLP)  guidelines. These blanks included trip
blanks, .equipment  blanks,  ambient condition  blanks,  laboratory calibration
blanks, and laboratory method blanks.  The dBase program performed two tasks in
the evaluation of quality control blanks:

      A.       Quality control blanks were analyzed for  the  presence of common
              laboratory contaminants,  including acetone, 2-butanone, toluene,
              methylene chloride,  and phthalate esters.  Contaminant detections
            •  were  considered valid only if the analyte concentration exceeded
              10  times  the  maximum  analyte  concentration  in  any •blanks
              associated with the  sample batch.  If an  analyte concentration did
              not exceed 10  times the maximum concentration for  an analyte found
              in any blanks, the  dBase program  inserted a 'B'  in the NDBLANKC
              field of the BCHRES.DBF analyte  record.  If  a 'B' was assigned to
              the NDBLANKC field,  the analyte concentration was not considered
              valid and was  deleted from further consideration.

      B.       For any other contaminant,  an analyte  detection was considered
              valid only if the analyte  concentration exceeded five times the
              maximum analyte concentration detected  in any blanks associated
              with  the sample batch.  If an analyte concentration did not exceed
              five   times the  maximum  analyte  concentration in any  blanks
              associated with the  sample batch, the dBase program inserted a 'B'
              in the NDBLANKC field of the BCHRES.DBF  analyte record.  If a 'B'
              was assigned to the NDBLANKC field, the sample concentration was
              not considered valid and was deleted from further consideration.
      3.      Laboratory Holding Time Analysis.   Laboratory holding times are
time constraints for  stages  in laboratory sample analysis. Laboratory holding
times were  evaluated to  identify all  samples  which exceeded  any laboratory
holding time with respect to extraction and analysis. Analytical results based
on missed holding times were rejected from further consideration.

      To accomplish this task,  a dBase program was developed to evaluate missed
holding times by analyzing one of the IRPIMS-formatted databases, BCHTEST.DBF
(batch test database) .  In BCHTEST.DBF, each record contained a sample collection
date, an extraction date, and an analysis date.  The dBase program evaluated each

-------
record in  BCHTEST.DBF to determine  if the analysis  for a sample  exceeded a
laboratory holding time.  Each BCHTEST.DBF record was associated wxth a set of
records in BCHRES.DBF which contained  the  analytical results for  the  set of
chemicals  run for  the analysis.   For each  record  in  BCHRES.DBF  contaxnxng
anllytical  results  for an analysis  which exceeded a holding  time,  the dBase
^ogram inserted  the  qualifier  'OUT'  in the record's HTFLAG field,  Durxng  the
construction  of individual site  databases, any records qualified with 'OUT' ,n
the HTFLAG field  were considered unusable.

      A dBase program,  SITESTAT.PRG,  was  used to calculate  statistics required
for the selection of organics and inorganics of potential concern for each .it..
This program  was  run  separately on each individual site database  for the entxre
depth  sampled and  for a specific soil sample  depth-of-concern.  - A ^th'^
concern is defined  as the soil interval which a receptor could contact.   If two
receptors  could contact site soil at different depths  (e.g.,  a  current worker
contacting surface soil and a  future excavation worker  contacting subsurface
soil),  the greatest  depth-of-concern was used to generated  statistical  datau
Statistics  included:  the   maximum   and  minimum  concentratxon,  the  average
concentration,    the   95%  UCL   of  the  arithmetic  mean  of  the  analyte s
concentrations, the number of samples evaluated, and the number of Action, for
 an analyte.  ^ Report Writer was used to print the  calculated statxstxcs from
 Se statistics  databases produced  by SITESTAT.PRG and an  additional  dBase
 program,  ABOVEMAX.PRG, in table form for each site.

       The  criteria for retaining or  deleting  inorganic constituents  involved
 comparing  inorganic  concentrations  at  each site to inorganic  concentrations at
 "ted background site. To accomplish this,  output from SITESTAT.PRG was
 evaluated using  the  following  approach.   First,  SITESTAT.PRG was executed for
 each site database for the entire depth sampled.  An inorganxc chemxcal whxch was
 not detected in  any  site samples was deleted from further consideration for the
 site     Second,  comparisons  were  made  using  the  maximum  concentratxon and
 arithmetic mean  for  each inorganic  chemical  at the background site and at each
 individual site for all collected  samples.   Table III provides an example  of
 statistics  calculated by  SITESTAT.PRG for  inorganics  (i.e.,  arsenxc  and
 magnesium)  for  a site.    Third,  the frequency  of  detections above maximum
 background concentrations for each  inorganic chemical within a sample depth-of^
 concern  was evaluated.   This frequency was obtained by  usxng  output  from
 SITESTAT.PRG and ABOVEMAX.PRO.  Specifically, SITESTAT.PRG identxfxed the number
  of samples analyzed  for each inorganic chemical within a sample depth-of-concern.
 ABOVEMAX.PRG identified the detections above maximum background concentratxons
  O&tf.iw

-------
: •.•.',;':•„;,: ...',;. . . sable iix
; Statistics Data
PARLABEL
MAX
MAX
SITEID
MAX
UNITS
MAX
SBD
MAX
SED
MAX
QUAL
AR
MEAN
ARMUCL
95
SP
SIZE
DETECTS
Inorganics
Arsenic
Magnesium
14.4
21765
1
1
mg/kg
mg/kg
10.0
10.5
10.5
11.0


5.1
8083
5.2
8623
173
158
2
158
Organics
Bis (2 -ethyl -
hexypthalate)
Xylenes
5.5
0.108
1
1
mg/kg
mg/kg
11.0
10.5
11.5
11.0


0.7
0.003
0.8
0.004
108
191
37
5
Key:
      PARLABEL
      MAX
      MAXSITEID

      MAXUNITS
      MAXSBD & MAXSED
      MAXQUAL

      ARMEAN
      ARMCJCL95
      SPSIZE
      DETECTS
Analyte label
Maximum concentration
Number designating  the site from which  the maximum sample was
obtained.
Units of MAX
Max's sample beginning depth and ending depth, respectively
Qualifier for the maximum concentration; if the MAX value was a
nondetect, an 'ND' would appear in this field
Arithmetic mean
95% UCL of the ARMEAN
Number of samples analyzed
Number of valid detections
Note:  This table was created from a SITESTAT.PRG database.

-------
for each inorganic chemical for all depths.  The frequency of detections above
maximum background concentrations  for each  inorganic  chemical was  manually
calculated  by dividing  the  number of  detections  above maximum  background
concentrations within the depth-of-concern by the total number of samples wichin
the  depth-of-concern.   Fourth,  the  magnitude  of  detections above  maximum
background concentrations within the depth-of-concern was evaluated using output
from ABOVEMAX.PRG.  Table IV provides  an example of magnesium detections above
maximum background concentrations for  a site.

      For  selection  of  organic  chemicals  of potential  concern,  a  similar
evaluation process was utilized  with the aid of SITESTAT.PRG output.   For organic
chemicals,  however,  the  background  site  was  not used  as a.   comparison.
Consequently, ABOVEMAX.PRG  was  not  used.   The primary analysis was made using
frequency  and  magnitude  of detections.    First,  as   previously  described,
SITESTAT.PRG  was  executed for the entire depth sampled.   All organic compounds
which were  detected  in less than  5  percent of the site  samples were eliminated
from further consideration.  Second,  SITESTAT.PRG was  executed for each  site
database for the depth-of-concern.   The frequency of detections within the depth-
of-concern  was evaluated.   Third,  dBase was  used to construct a  database  of
detections  for all  depths.   The magnitude  of each detection within  the  site
depth-of-concern was evaluated.  In general,  compounds  which were not  detected
at  levels  significantly  greater  than  the  laboratory  detection  limit  were
 eliminated from  further consideration for the site.   Tables 3 and 4 provide
 examples of  SITESTAT.PRG  results  used to justify  retention or  deletion  of
 organics of potential concern.

       Table V provides an example of justification for retention or deletion of
 inorganic and organic chemicals for a site, using criteria previously discussed.

       Figures 1  and 2 provide  a  general description of the steps performed by
 SITESTAT.PRG. It is  beyond the scope  of this  paper to provide a detailed
                                        10

-------
Table IV
Detection By Depth
Inorganics1 :
(only concentrations above the background level are given)
Site
ID
1
1
1
1
PAR
LABEL
Mg
. « Mg
Mg
Mg
LOCXREF
MWA1
MWA1
MWA1
MWA1
ANMCODE
SW6010
SW6010
SW6010
SW6010
SBD
10.5
90.5
120.5
215.5
SED
11.0
91.0
121.0
216.0
PARVQ
=
=
s
=
LABDL
50
50
50
50
PARVAL-
DLOM
21,765
16,230
16,216
18,003
Organics2
Site
ID
1
1
1
1
1
PAR
LABEL
Xylenes
Xylenes
Xylenes
Xylenes
Xylenes
LOCXREP
MWA1
MWA1
MWA1
MWA1
MWA1 .
ANMCODE
SW8240
SW8240
SW8240
SW8240
SW8240
SBD
10.5
70.0
80.5
90.0
100.0
SED
11.0
70.5
81.0
90.5
100.5
PARVQ
=
=
=
=
=
LABDL
0.005
0.005
0.005
0.005
0.005
PARVAL-
DLUM
0.108
0.018
0.013
0.027
0.017
Mg
Xylenes
Magnesium
Total Xylenes
'inorganic'data table was created from an ABOVEMAX.PRG database.
Organic  data table  was created using dBase.
                                      11

-------
Inorganics
  Silver
    (Ag)
  Arsenic
    (AS)
 Magnesium
    (Mg)
  Mercury
    (Hg)
    Zinc
    (Zn)
   Bis (2-
 ethylhexyl)
  phthalate
    (BEHP)
Chlorobenzene
Total Xylenes
               .  .       Table V
       Examples of Justification for Selectionc-i
        Chemicals ot Potential Conee«i at a Site
	Explanation
Ag is  detected in only  l  of 131 sample

                                                                    mean.
                                        f
for  the  site is less  than  the background
concentrations are considered to be at background
Only 4  of  158  sample  concentrations  exceed  the
background level.  Three of  the samples were taken from so:.!
moan nnlv sllcrntJ-V  exceeue  unc  uo.v-r>.aj.——-—	
C^entrations9arey considered to be within background at the
soil depth-of-concern.		
          sample  r-nnrAntrations  are  below
          limit.
All  site
detection
background levels.
      of
concentrations  are
  Concentrations  ar
                                             the  laboratory
                                       considered to  be at
                                 . -        - - - — : - --
 Five  of  158  sample  concentrations   exceed   the  maximum
 background level ^Pour of these  samples were taken from soil
           ensure concern (146 mg /kg wfll 375 mg/kg  at  0 0
 BEHwas detected in 37 o, : 10 e » sa^ e^

 signriic^tly" "exceed t'he laboratory detection limit.  Because
 of  the  sionificant  concentrations  in  the soil  deptn-ot
 concern, alid because of the  relatively high ^^/^ection
 frequency  (34%),  this   analyte   is- retained  for  turther
 cons ideration .
 CUHtjJnacJ.&i—IWA* .	_^	                 . —,*«/•77

 sjsnssr-r. trr??Jf, •«-. SUSE^-S^ s
 5%    Three of  the  detections are very  low,  only
 exceeding the  detection  linit.   Consequently,  the
 detection frequency  is  4*.   Chlorobenzene  is  deleted from
 further consideration.		_	.	
 Total  xylenes  were  detected in 5 of  191 samples (0.013  to
 0 0108 mg/kg at 10.5  to  100.0 feet).  Because of low sample
 detection^  frequency  (3*)  and because   there  is only  one
 detlctio^ JlttSTtS soil depth:of-concern,  total xylenes is
 deleted from further consideration.	,	
                                                                            Status
                                                                            Deleted
                                                                            Deleted
                                                                            Deleted
                                                                            Deleted
                                                                            Retained
                                                                            Retained
                                                                             Deleted
                                                                             Deleted
     OZ&n*
                                          12

-------
                              (   BEGIN   J
                             Set the Soil Interval
                              Depth erf Concern
                               'MAIN LOOP?
                                Process Site
                             ' Database RecordsN
                             v Sequentially Until y
                                 the End of
                                 Database
                         Initialize Statistical Variables.
                         Identify Current Chemical to
                               be Processed.
                     Delete all  records from the working
                     database   UCLFILE.DBF.   Chemical
                     sample  concentrations are placed in
                     UCLFILE.DBF to calculate the 95% UCL
                     of    arithmetic    mean    of    the
                     concentrations.
                               INNER LOOP: \
                               Process Each   \
                             Record Sequentially  \
                               for the Current     J
                              Chemical Being   /
                                Processed    /
                                Concentration
                                 Subroutine
                          I   Skip to Next Record   |
                          j     End Inner Loop     j
                                 Statistics.
                                 Subrouting
                         \     End Main Loop     |
' 7h* E*rth Tvchnolcgy
 Corporation
                                                                                           FIGURE 1
                                                                  SITESTAT.PRG FLOW DIAGRAM
•0236.FK3
                                                             13

-------
    (    BEGIN
      Concentration*
    I   Subroutine
3
                                        Sample Concentration -
                                   Laboratory Detected Concentration.
                                     Increment Number of Detections
                                              by One.
   Sample Concentration -
   Lab Detection Limit -0.5
        Add Sample
   Concentration to Sum for
      Arithmetic Moan
    Increment Number of
    Samples Processed by
           One
       Append Sample
       Concentration to
      Working Database
       UCLFILE.DBF
             Hon GraaUr
        thtn *• Working
      Uudmwn U»InUJmd (or
         ihtCtwmlcJl?
      Conanlralfon l**» timn
       «w Working Minimum
                  Reset Working Maximum to the
                  Sample Concentration. Reset
                  Working Sample Parameters
                  for the Maximum including:
                  LOCXREF. SACODE,
                  ANMCODE. EXMCODE, SBD,
                  SED, RAQUALIFY
                 Reset Working Minimum to the
                 Sample Concentration. Reset
                 Working Sample Parameters for
                 the Minimum including:
                 LOCXREF. SACODE.
                 ANMCODE. EXMCODE. SBO,
                 SED. RAQUAUFY
                                                                       BEGIN
                                                                      Statistic*
                                                                     Subroutine
                                                                               Calculate  Arithmetic  Mean  -
                                                                               Sum of Sample Concentrations /
                                                                               Number of Samples
                                                                               Open Working Database
                                                                               UCLFILE.DBF. Use the Recorded
                                                                               Sample Concentrations to
                                                                               Calculate the Standard Deviation
                                                                               for the Concentrations.
                                                             Obtain  t-value  from  Database
                                                             TVALUE.DBF.   Calculate  95%
                                                             UCL of the Arithmetic Mean of the
                                                             sample concentrations.
Record Statistics for the Chemical in
Database SITESTAT.DBF.  Record
the soil depth-of-concem.  Record
the    Maximum   and  Minimum
Chemical Concentration and  the
Associated Parameters. Record the
Number  of  Samples, Number of
Detects.  Arithmetic Mean. 95% UCL
of  Arithmetic  Mean,  and  the
Standard Deviation for the  Sample
Concentrations.
                                                                                     f    END   ^
                                                                                          Statistic*   I
                                                                                     I   Subroutine  J
            END
        Concentration
          Subroutine
  D
                                                               1 ThiEfrth Technology
                                                                Corporation
                                                                        FIGURE 2
                                                                  SITESTAT.PRG FLOW DIAGRAM
                                                                             SUBROUTINES
KOOS.FK3

-------
discussion of SITESTAT.PRG. Note that the program requires creation of an index
file for analysis of a site database.  The index file must group all samples for
an analyte in ascending order according  to the sample beginning depth. This  is
accomplished  by  indexing  the  site  database using  the  concatenated  fields
PARLABEL+SBD.

IDENTIFICATION OF HUMAN RECEPTORS

      Conceptual site models were developed for each site.  Each conceptual site
model  includes  a  description of  previous  or  current  activities,   including
chemical sources and amounts', identification of chemicals of potential concern
which can be attributed to site activities, identification of migration pathways,
and identification of current or  future human receptors who could be impacted  by
contact with contaminated site soil or chemicals migrating from a  site.

      Three migration pathways were  identified  whereby chemicals of  potential
concern  could  potentially  migrate  toward human  receptors.    These  migration
pathways are:

      •     Migration  of  volatile  .organic  compounds  (VOCs)   from  the soil
            saturated and unsaturated zone -to air

      •     Migration of dust containing particulate-bound contaminants to air

      •     Transport of surface soil contaminants  by ephemeral surface water
            runoff.

      Human receptors who could be impacted by migration of site contaminants or
by direct contact with site  soil were identified.   These  receptors are:  current
workers, current offsite resident children, and future excavation workers.
                                      15

-------
      First, current workers were identified who could contact contaminated site
soil directly.  These receptors were workers in buildings located on sites with
contaminated soil or workers  whose  activities brought them in contact with site
soil, such  as  workers who jog across a site with  contaminated soil.   Second,
onsite current workers were identified who could be impacted by migration of VOCs
from the site  soil  saturated and unsaturated zone or migration of contaminated
dust  to air.   Third, offsite  current workers were identified who  would be
maximally impacted  by VOCs migrating  off  site  from soil or by dust containing
particulate-bound contaminants migrating off site.   Fourth,  current  resident
children  (i.e., 6 years of age)  were identified as a sensitive subpopulation who
could be  impacted by  contaminants migrating from  a site in ephemeral surface
water runoff through drainage  channels and ditches.  Finally,  future  receptors
were  identified who could  be  impacted  by  onsite  excavation  activities.
investigations revealed  that excavation activities could occur in the future  at
three sites.  Therefore,  future excavation workers were identified at these sites
who could be impacted by airborne VOCs or contaminant-bound dust,  or who could
 contact the contaminated soil directly.

 IDENTIFICATION OF EXPOSURE PATHWAYS

       Exposure pathways were identified for human receptors who could be exposed
 to a chemical through a  migration pathway or by direct contact  with the chemical
 in  soil.

       The  following  exposure  pathways   were  identified  for current  worker
 receptors  who could be  impacted by soil migration pathways or  who could  contact
 soil directly:

        •      incidental  ingestion of  contaminated surface  soil
        •      Dermal absorption of chemicals from surface soil
        •      Inhalation of  contaminated dust
                                        16

-------
      •     Inhalation of VOCs.

      Exposure pathways for offsite resident children who could contact chemicals
in soil transported from a site in the drainage channels and ditches are:

      •'    Incidental ingestion of contaminated surface soil
      »     Dermal absorption of chemicals from surface soil.

      The existence of surface  water in the drainage  channels'and ditches is
temporary  (i.e.,  on  the order  of hours),  and only  11  inches  per  year of
precipitation  falls at the facility.   Consequently,  the  exposure  of  offsite
children to contaminated surface water is negligible, and resultant  surface water
exposure pathways are considered to be incomplete.

      The following  exposure pathways were identified for  future excavation
workers:

      •     Incidental ingestion of contaminated subsurface soil
      •     Dermal absorption of chemicals from subsurface soil
      •     Inhalation of contaminated dust
      •     Inhalation of VOCs.

DETERMINATION OF EXPOSURE CONCENTRATIONS AT RECEPTORS

      Both the arithmetic mean and 95% UCL of  the arithmetic mean contaminant
concentrations at  receptors were  quantified per  federal  and  regional  USEPA
guidance. Because data were not collected in an unbiased manner and goodness-of-
fit statistical tests  rejected log-normal distribution of a  subset  of soil data,
data  were  assumed  to  be  normally  distributed.    Consequently,  arithmetic
concentrations, rather J:han geometric concentrations, were quantified at receptor
intake locations.
                                      17

-------
      .Table VI presents the formulas used to quantify the arithmetic mean and 95%
 UCL concentration at a receptor for each chemical of potential concern:

 Soil Contaminants

                                       ^ *.  _=i ™,iat-P 'site soil mean and 95% UCL
       The SITESTAT.PRG program was used to calculate  site
 exposure concentrations at receptors for varying exposure depths - of - concern.  As
 stated previously, if a 'B< was assigned to the NDBLANKC field or a. 'OUT'  was
 assigned to  the  HTFLAG field,  reported analyte concentrations were considered
 unusable.  Nondetectable concentrations were represented by an •»>'  in the PARVQ
 field of BCHRES.DBF.  In accordance with USEPA guidance, for  all concentrations
 of  chemicals  of  potential  concern  which  were  reported   as  nondetec.able
 concentrations  (i.e., the chemical concentration does not  exceed the laboratory
 detection  limit) , a value of one-half the laboratory detection limit was used for
 calculating  exposure concentrations  by SITESTAT.PRG  (3)  .   All  valid  analyte
 detections were  represented by '=' in  the PARVQ field in BCHRES.DBF.   If an '='
 was assigned to the  PARVQ field, the  dry weight  soil  concentration in  the
 PARVALDLUN field was used by SITESTAT.PRG to calculate exposure concentrations.
  in those cases  where the  95% UCL  of the mean concentration exceeded the maximum
  detected chemical concentration,  the  maximum  detected concentration was used as
  che exposure concentration.   Examples of arithmetic mean and 95%  UCL exposure
  concentrations have been presented in Table III.

        For current worker soil pathway exposures, sample data collected from the
  surface interval were used to  calculate surface soil concentrations.  For future
  excavation  worker  exposure,   sample  data collected   from  assumed subsurface
  excavation  depths (e,g., 20 feet depths for two sites,  10  feet depth for one
' site) were  used to calculate  soil exposure concentrations.

        For the current resident child exposure, the resident  child was assumed to
  be exposed  directly to chemical  concentrations detected  in  site  surface  soil.
   OZ3«,rev
                                         18

-------
   ••.. •  ••"••'•'•'•'•.. ';<:•'••;':"••-': .••'v-:-y!-;.r.-?V; -• Table VI    ''••..  . " :  • .-•  .
   Formulas : Used tb; Calculate the ..Arithmetic Mean, and 95% UCI*
     of the Arithmetic Mean Concentrations at Human Receptors
                          Arithmetic Mean:
where :
        x
        n
        Q
                          The arithmetic mean concentration
                          Number  of  contaminant samples
                          Contaminant concentration.
                              95% UCL:
             95% UCL of the arithmetic mean = x
                                                „_,)
 where:  x      =   The arithmetic mean concentration
         fco.95.D-I  =   T*16 9s*  fc distribution value for n-1 'degrees  of
                    freedom
         s      =   Standard deviation
	n	=   Number of contaminant samples.	
                                 19

-------
It was conservatively assumed  that  site  chemicals  were totally transported to
resident child receptors by surface  water runoff, without any chemical dilution
(eg., adsorption to  soil).    Consequently,  the  site surface  soil chenucal
                                       j  *.   v,o  t-he  surface  soil  exposure
concentrations  calculated  are  assumed  to  be  the  surra<-«
concentrations for the resident child.

Air Contaminants

      Soil gas data were used to calculate VOC concentrations in air for current
worker and future excavation  worker  exposure.   The Farmer Model was used to
determine  emission  rates  of  VOCs  from  each  site   to  determine exposure
concentrations for current workers.  The  Farmer Model is a modified Fick's First
Law for steady-state diffusion. Fick's First Law assumes that transport of a VOC
through the  soil  cover layer is controlled by molecular dn.ffus.on.  It does not
account  for the  effects  of  atmospheric  conditions,  such as temperature,  wxnd
speed, and barometric pressure, upon  emission rates.

       Using chemical air diffusion coefficients for each VOC  and assuming the
 total soil porosity and air-filled soil porosity of the site-specific .oil to be
 30% and 10%, respectively,  the Farmer Model was applied to the average soil gas
 concentration and  95% UCL  of each  compound.    The  VOC  emission  rates  were
 determined in units of milligrams per square  meter-second  (mg/m*-s) .  The total
 mass of VOCs emitted per second was determined by multiplying the VOC emission
 rate by the area (in square meters) of each site.

       To determine the ambient air concentration of VOCs for onsite or offsite
 worker receptors, the  Industrial Source Complex-Long Term (ISC-LT) dispersion
 model program  was  used.   The ISC-LT is a USEPA-approved Gaussian dispersion
 model.  The model operates  in both long-term and short-term modes.  The model
 uses meteorological data, including wind speed,  wind direction, and atmospheric
 stability class, and area or  point-source chemical airborne concentrations,  to
 determine impact at a receptor location. The model program was used in the area
 source and  long-term modes to  estimate worst-case airborne VOC concentrations at
 maximally impacted worker receptors.   The location of the nearest  (onsite  or
 offsite)  suitable  receptor was used  to  determine  the  maximum  exposure for that
 site.

       The ISC-LT dispersion model  was used to estimate exposure concentrations
  for offsite,  maximally exposed  workers.  These  workers were   identified  by
  selecting the building at which the model predicted a maximum concentration.  The
  ISC-LT dispersion model was  also  used for onsite worker receptors because it
  predicts long-term exposure  concentrations  needed to quantify  chronic worker
  023&IW
                                        20

-------
exposure.

      The average and 95% UCL onsite or offsite worker exposure concentrations,
in milligrams of VOC per cubic meter  of  air, were  determined for each detected
voc.

      For future exposure,  an assumption was made that all VOCs  in the soil would
be released to the atmospheric excavation volume.   The VOC  concentration in  the
air per hour can be calculated using the box model method.  The box model method
assumes steady-state, hourly emission  rates and uniform dispersion conditions so
that VOC emissions are uniformly  distributed throughout a "box" which is defined
by the area of the  source and the mixing height.  The box model is.applicable  for
estimating airborne VOC exposure concentrations for excavation workers  because
it uses short-term conditions (e.g.,  average annual windspeed,  hourly emission
rates) to estimate exposure concentrations for  short-term  excavation  tasks.

      Applying the box model method to the average  and 95%  UCL concentration of
VOCs in the soil gas yielded the average  and 95% UCL  air concentration of VOCs
which may be inhaled by excavation workers.

      Inhalation of contaminated dust by current workers was  also investigated.
Particulate matter  (PM-10)  data for the site vicinity were obtained,  including
the 24-hour maximum concentrations and the quarterly averages for 1990  and 1991.

      The average and 95% UCL concentrations  for chemicals  of potential  concern
were determined from surface soil sampling results. Chemical concentrations in
airborne  dust  were  calculated  by  multiplying   the chemical  surface  soil
concentration by the  appropriate PM-10 air concentration.   The average PM-10
value and an average surface soil concentration  were used to determine an average
dust  concentration.   The  maximum PM-10  value and  the  95% UCL  surface  soil
concentration were used to determine  95%  UCL dust  concentrations.

      For future exposure pathways, dust concentrations  inhaled by excavati'on
workers were  determined  using similar parameters  and assumptions used in  the
determination  of  VOC exposure   concentrations.    Using  assumed excavation
parameters, a particulate emission rate for backhoe excavation work was obtained
from the USEPA Air/Superfund  National Technical Guidance Study Series, Volume
III,  January, 1989 (5).
OZM.rev
                                      21

-------
                                           •„„ t-vio  box model method, the total
      Mt.r determine . mixing volume u.=.ng the ^» «o                 l,te
»,ount (in ..trie tons,  of soil to  be removed .as multWl"  W
                          in air
to the calculated concentrations of
concentrations of chemicals which may
ESTIMATION OF DAILY INTAKE VALUES
                                      be  inhaled by excavation workers.
       intake values were  estimated for identified  exposure pathways    Human
 intake (i.e.,  the magnitude of exposure) is expressed as the amount of dta-xcal
 Tan exchange boundary  (e.g.,  skin, lungs,  gut)  which  is  ava,   , le  f r
 GDIs were  estimated for an exposure  of 7 years  to a lifetime  and SDIs were
 estimated for exposure of two weeks to 7 years  (3) .

       The basic formula used to estimate CDI or SDI is presented in> Table VII.

       As previously described, SITESTAT.PRG was used to calculate the mean and
 95% UCL  of the mean soil chemical  concentrations at receptors *«^-£-
 These  data  were  established  in  statistical  databases  by  SITESTAT^
 Concentrations of VOCs in air at receptors were estimated using the  Farmer Model
 and air  dispersion modeling.   «-lO data and surface soil  concentrations were
 "ed to determine dust concentrations at receptors.  The arithmetic mean chemical
 concentration was used to quantify average intake; the 95% UCL of the arithmetic
 "emical concentration was used to o^antify the reasonable maximum ensure
             accordance  with USEPA  (3) .   As stated  previously,  if  the 95% UCL
                exceeded the maximum concentration for a chemical,  the maximum
  concentration was .used.

        Standard default exposure factors  were used  to estimate  intake .where
  applicable  (3) (4) ; reasonable assumptions were made to quantify site-specific
  exposure factors.

        The  exposure frequency  for all permanent  current  workers  and  future
  excavation workers was assumed to be 250 days/year (4) .  The assumption  is made
  that an adult is at work 5 days/week  for 50 weeks/years.   Workers  were  assumed
  OZ3«.rev
                                         22

-------
          ,,..  .....,...     .                 ...
           Biasic Formula Used to Bstiifiater Intake. Values
              CD/ or SDI (mg/kg-day) = C x
CR x EF x ED
    BW
J_
AT
 Where:

 GDI or SDI   =  GDI or SDI by the receptor  in mg/kg body weight-dry
    C         =  Chemical concentration; the arithmetic mean or 95%
                 UCL of the mean concentration contacted over the
                 exposure period
    CR        =  Contact Rate; the amount of contaminated media
                 contacted per unit time or  event
    EF        =  Exposure Frequency  (days/year)
    ED        =  Exposure Duration  (years)
    BW        =  Body Weight of receptor; the average body weight over
                 the exposure period  (kg)
    AT        =  Averaging Time; period over which the exposure is
	a.veraged (days) .	
                                23

-------
to jog on a contaminated site for 30 minutes/day, 5 days/week,  and 50 weeks/year.

      Resident children were assumed to play off site in the drainage channel for
4 hours/day  (exposure  time) at 1 day/week  for 50  weeks/year,  or 50 days/year
 (exposure frequency).

      The exposure duration for future excavation at two sites was assumed to be
                                          j ,.„*-•! ^n -For future excavation at the
 150 hours (i.e., 0.075 year); the exposure duratxon for rutu
 third site was  assumed to be 500 hours (i.e.,  0.25 year).

      Two  dBase  programs,   AIREXPS.PRG  and  SOILEXPS.PRG,   were  developed to'
 estima" average exposure and RME for identified  air and soil ensure^athways
 respectively.   These  programs  implemented standard default and  site  specific
 exposure factors for each identified exposure pathway.  For each site with an
 idlntlfiefvoc  or dust exposure pathway(s) ,  AIREXPS.PRG used  the arithmetic mean
 Ind  95% So, of  the mean  chemical  concentrations  for  VOCs  or  dust  at site
 receptor(s)  to quantify the average and RME for each chemical.

       For each site with an identified soil  exposure pathway(s), SOILEXPS.PRG
 used thl arctic mean and 35, UCL of  the mean chemical concentration at site
 receptor(s)  to quantify the average and RME for each chemical.

 CONCLUSIONS

       DBase was  used to manage  several IRPMIS-formatted databases containing
 analytical soTl data for several hazardous waste sites.   Several  dBase programs
 were developed to automate  portions of the risk assessment process, including the



 provided several advantages:

        .  Analytical  soil data were evaluated for consistency and completeness.
        .  A large quantity of data was evaluated within a short  timeframe.
        .  Accuracy  of  the  risk assessment  calculations was  improved.
        .  A large quantity of data was  evaluated  in  a cost  effective  manner.

  ACKNOWLEDGEMENTS

        Numerous Earth Technology personnel were involved in the  risk  assessment
  process.   In particular,  the  authors would like  to acknowledge ?^a««fl«.
  Pristine Pryately, Sandra Karcher, Natalie  Paul, Pamela Anderson,  and Dana Lynn

                                         24

-------
Bowers for their creative and diligent efforts.

REFERENCES

(1)    U.S. Environmental Protection Agency,  Region I Laboratory Data Validation
      -- Functional Guidelines for Evaluating Inorganics Analyses. Washington,
      D.C., June, 1988.

(2)    U.S. Environmental Protection Agency,  Region I Laboratory Data Validation
      -- Functional Guidelines  for Evaluating Oraanics  Analyses. Washington,
      D.C.,' February,  1988.

(3)    U.S.  Environmental  Protection Agency,  Risk  Assessment  Guidance  for
      Superfund,  Volume  I:  Human Health  Evaluation  Manual  (Part  A).  Interim
      Final,  Washington, D.C.,  December,  1989.

(4)    U.S.  Environmental  Protection Agency,  Risk  Assessment  Guidance  for
      Superfund.  Volume I; Human Health Evaluation Manual Supplemental Guidance,
      "Standard Default  Exposure Factors".  Interim . Final,  Washington,  D.C.,
      March,  1989.

(5)    U.S.  Environmental  Protection Agency/ Air/Superfund  National  Technical
      Guidance Study Series, Volume III,  Research Triangle  Park,  NC,  January
      1989.
                                      25

-------
        BIOGRAPHICAL SKETCH: DR. JAHUSX Z. BYCZK0WSKX

Dr. Byczkowski is currently a Project scientist and Study
Director at ManTech Environmental Technology, Inc. in
Dayton, OH. He received his Master's Degree in Toxicology
from the School of Pharmacy, Academy of Medicine in GdansK,
Poland in 1970. His Ph.D. in Pharmacology and D.Sc. in
Biochemical Pharmacology were received from Academy of
Medicine in Gdansk, Poland in 1975 and 1979, respectively.
He holds a professional license of Pharmacist since 1970.

Prior to joining ManTech in 1991, Dr. Byczkowski held a
tenured position at the Academy of Medicine in Gdansk,
Poland, was a Cancer Research Scientist in Roswell Park
Memorial Institute in Buffalo, NY, was involved in research
and teaching in Toxicology Program at University of South
Florida, college of Public Health in Tampa, FL, and served
as consultant to Pharmaceutical Specialties, Inc. in Tampa,
FL.

Dr. Byczkowski has been the author/co-author of 57
publications in peer-reviewed journals, 3 book chapters,
made 50 presentations at meetings and was invited to make  12
presentations at different Universities. He  is a member  of
The Society of Toxicology, The  Oxygen Society, International
Society for Free Radical Research,  Society  for Research  of
Polyunsaturated Fatty Acids, American Association  for  the
Advancement of Science, American Museum of  Natural  History,
The New York Academy of  Sciences.

-------
       LACTATIONAL TRANSFER OF TETRACEELOROETHYLENE IN RATS

                    'Jamusz Z. Byczkowski and 2Jeffrey W. Fisher
               'ManTech Environmental Technology, Inc., Dayton, OH and
              2Toxicology Division, Occupational and Environmental Health
                      Directorate, Armstrong Lab., WPAFB, OH.
                Abbreviated title: LACTATIONAL TRANSFER OF PCE
                Send correspondence to: Janusz Z. Byczkowski, ManTech
             Environmental Technology, Inc., P.O. Box 31009, Dayton^ OH
                                   45437-0009.
ABSTRACT:
Tetrachloroethylene  (PCE) is a commonly used organic solvent and  a suspected human
carcinogen, reportedly transferred to human breast milk following inhalation exposure. Transfer
of PCE  to  milk may  represent  a threat to the nursing  infant.  A  physiologically-based
pharmacokinetic (PB-Pk) model was developed to quantitatively assess the transfer of inhaled
PCE into breast milk and the  consequent exposure  of the nursing  infant. The  model was
validated in  lactating rats. The  model  described the distribution  of inhaled PCE in maternal
blood and milk as well  as the nursed pup's gastrointestinal tract, blood and tissue. Lactating
Sprague-Dawley rats were exposed to PCE at concentrations ranging from 20 ppm to 1000 ppm,
via the inhalation and then returned to their nursing, 10-11 days old pups. PCE concentrations
in the air, blood, milk and tissue were determined by gas chromatography (GC) and  compared
to model predictions. Several prognosis for kinetics of PCE  distribution in exhaled air, blood
and milk of exposed  human subjects were run and compared with  limited available human data
from  the literature.  It is concluded that the PB-Pk model described fairly accurately the
concentration of PCE in both lactating rats and humans.
KEY WORDS: TETRACHLOROETHYLENE, PERCHLOROETHYLENE, BREAST MILK,
                              RAT, HUMAN INFANT.
1. INTRODUCTION
    Transfer of toxic chemicals via mother's milk represents an important, although not widely
recognized health risk to the infant. An evaluation of the  hazards of exposure to occupational
chemicals  transferred  from  mother to  baby must  include  qualitative and  quantitative
determinations of chemicals that contaminate breast milk.
    A review of the literature conducted by  Cone et al.  (1) for the USA  Environmental
Protection Agency revealed that many chemical compounds may be transferred with breast milk
during feeding.  Most of these chemicals were either environmental pollutants or drugs. An
extensive list of  the chemicals detected or excreted in human milk (about 150 compounds) has

-------
been published recently by Giroux et al (2). The lactational transfer of both environmental
oollutants and drugs was reviewed in two recent publications (3, 4).
    Volatile organic solvents deserve special attention because, these chemicals are widely used
in industrial facilities. Inhaled volatile organic chemicals quickly transfer to systemic circulation
where they  selectively  partition into fat stores, including breast milk.  The residence time for
volatile organic chemicals in the body (including breast milk) is not long when compared to
rahfent environmental contaminants such as polychlorinated biphenyls, but die levels achieved
Ke fat stores such as breast milk may be substantial. For instance, tetrachloroethylene  syn.
Perchloroethylene or PCE) was detected in  milk from one Canadian woman who  regularly
vTsUedTr husband during his lunch hour at  a dry-cleaning factory (5). ^™**?«%*
PCE  in breast milk was 10 ppm one hour after the visit and over the next 24 ^hours fe » PCE
concentration in breast milk decreased to 3 ppm. Her infant developed Jaundice at Ae age of 6
weeks, but recovered  quickly  after cessation of breast-feeding (5).  While the disease was
attributed to the contamination of breast milk with PCE, it is not clear whether the association
with  obstructive jaundice was casual or spurious.                            ...      < •
     PCE is a volatile, nonflammable liquid widely used in the dry cleaning industry and in
metal degreasing operations. Acute inhalation of PCE vapor by humans has produced central
nervous system depression ranging from lightheadedness and muscular incoordmation at low
concentrations, to loss of consciousness and respiratory paralysis at higher doses (6-9). The
development  of minor, reversible hepatic dysfunction several days following accidental human
exposure to anesthetic concentrations of PCE have been noted also (6,7).
      There was limited information in the literature on tissue concentrations of PCE in rats or
 mice resulting from test exposures from which a PB-Pk model describing  the pharmacokinetics
 of PCE was developed  and  validated  (10). But  there  was no  information  available on
 pharmacokinetics of PCE excretion with milk.
      In this report a PB-Pk model for lactational transfer of PCE is described and validated in
 nursing rats. Several computer simulations and prognoses for the long-term PCE distribution . in
 exhaled air, blood and milk of exposed human subjects were done and compared with available
 human data from the literature. The computer simulation of the kinetics of lactational transfer
 of PCE may aid a quantitative assessment of the dose passed by the  exposed mother to the
 nursing infant.

 2. MATERIAL AND METHODS


  * * The* amount of PCE metabolized by animals in a closed gas uptake chamber (7.9  L) was
  measured by gas chromatographic  analysis as described by Gargas et aLJJLl). The amount of
  PCBtathe sampled air was measured by the HP  5890 Series E GC with FID detector. For each
  gas  uptake run, three lactating female Sprague-Dawley rats were used.
      The decrease in  PCE chamber concentration is indicative of the rate of metabolism of Ifte
  chemical by the animal (11). Analogous to  the description of kinetics for isolated enzymes by
  Selis and Menten theory, a "pseudo VMAX" and an "apparent KM" were determined along
  with a first order "metabolic rate constant,  KFC" by PB-Pk modeling.
  2.2. Blood and Milk Analysis                                            .         ,
      Samples of 80 pL of blood or milk were collected in triplicate from each animal into glass

-------
capillary tubes and then transferred directly to autosampler vials and extracted using n-hexane.
The extracts were analyzed by gas chromatograph equipped with a Vocol™ fused silica column
and an electron capture detector. Calibration curves were prepared and evaluated statistically for
the best fit.  Concentrations in  blood and milk  were corrected for  appropriate extraction
efficiency, determined by spiking blood and  milk with PCE  (96.2%±1.5 and 95.1%.+1.4,
respectively; n=9). The standards were processed also with each series of samples.

2.3. Tissue Analysis
     Tissues of pups euthanitized with CO2 and bled, were either placed in sample bags and
frozen in liquid N2 or placed in jars with n-hexane and processed fresh. Thawed or fresh samples
were homogenized and extracted with  n-hexane. The extracts were analyzed by GC analogous
to blood and milk extracts. The difference between extraction efficiency calculated for frozen
tissues (56.13%Ji6; n=4) and fresh tissues (56.73%±9; n=3) was insignificant.  Calibration
curves were prepared  using tissues of control pups spiked with known amounts of PCE.

2.4. Determination of Partition Coefficients
     A smear method was used for determination of tissue, milk and blood partition coefficients.
The fresh or frozen tissues were homogenized and about 0.1 g of muscle, liver, kidney, pup,
or 0.05 g of adipose tissue, or 250 pL of blood or milk were smeared on the walls of tared, 25
mL, borosilicate glass scintillation vials. The vials were weighed, sealed and then injected with
known amounts of PCE from an equilibrated standard bag. The vials were then incubated with
vortexing for 3 h at 37°C. Aliquots of 1 mL of head space were injected automatically onto the
gas chromatograph and analyzed as described above for blood and milk. The blood/air and
tissue/air partition coefficients were then calculated,  essentially as described by Gargas et al.
(12).

2.5. Animal Exposure
     Lactatihg female Sprague-Dawley  rats (body weight 232 - 352 g) were used as test animals.
After delivery litters, were reduced to  8 pups per dam and kept undisturbed for 10 to 11 days
(body weight of pups 16.2 - 27.9 g). On day 10 or 11 post portion, the lactating females were
exposed to PCE either in the closed gas uptake chamber (3 rats per 7.9 L chamber) for up to
6 hours or in an open inhalation chamber (5 rats per 30.0 L chamber) for  1 to 6 hours. Numbers
of rats  included in the gas uptake and in constant concentration exposures are shown in captions
to figures (typically, 5 groups of 5 dams).
     Inhalation was selected as the route of administration most relevant to occupational exposure
of women. Concentrations for inhalation exposures of rats were set between 20 and 1000 ppm
for 1 to 6 hr. At a selected exposure level (600 ppm) the time-dependent, measurements were
made in dam milk and blood after 1, 2,  3, 4, and 5 h exposure to PCE. Another group of dams,
exposed to 600 ppm of PCE for 2 h, was returned to their nursing pups, and blood as well as
tissue levels of PCE were measured in  pups at selected times after the maternal PCE exposure.
    The animals used in this study were handled in accordance with the principles stated in the
"Guide for the Care and Use of Laboratory Animals" prepared by the Committee on Care and
Use-of Laboratory Animal Resources,  National Research Council, Department of Health and
Human Services, National Institutes of Health, Publication No. 86-23, 1985; and the Animal
Welfare Act of 1966, as amended.

-------
                                in  S.MUSOLV, a Fortran-based contocus

                            ^
(VAX8530, Digital Equipment Corp., Maynard, MA).
3. RESULTS
(13). Additional compartments were added to describe milk (14) and ^WM.to*Uy>
for simplicity the pups were described by the lungs, arterial and venous blood, and the  other
tissue? ^^compartment  (16). However, this simplified model did not describe adequately the
        of PCE in  pup's blood and tissues. Milk was retained in the gastrointestinal ttac of
        pups which apparently delayed absorption of PCE for several hours. To describe this
            , ^additional compartment was added to simulate the gastrointestinal tract of pup

(Fig;  1).
                              equations describing each compartment building the PB-Pk
 model for lactational transfer of PCE (schematically shown in Fig. 1).
     For well stirred  compartments  without metabolism or other losses (fat tissue, slowly
 perfused and rapidly perfused tissues, pup tissues) the amount change over time is described as
 follows:
 Concentration in the tissue, Ci, is equal to AIM where Vi represents the volume of the i-th

 compartment).        ^ & ^  ^  (RAM)  is added to the weU

 description to account for metabolism (equal to VMAXC*CVL/(KM+CVL)+^*C   ^
 where VMAXC is pseudo-maximal  velocity  rate of  PCE - metabolism ^ CVL is  venous
 Concentration leaving the liver, KM is apparent Michaelis-Menten constant, KF is the first order
 rate of metabolism):

               dAL/dt = (QL(CA-CVL)-RAM)
      Quite analogous, for mammary glands compartment the ^^^'o
  contains a loss term for elimination for PCE from milk to pups, RPUP (equal to CMAT OU
  wheTcMAT is concentration in milk, OUTX is periodic zero order milk yield per dam).

               dAMAT/dt = (QMT(CA-CVMT)-KPUP)

  where QMT represents mammary blood flow, CVMT represents ve

                                           •4

-------
                       Oann
               ct
                                        OP
                  QC
              cv
                           Afoeoli
                          Lung Blood
•+-CX
                          pat Tissue
                        Slowly Perfused
                                        OF
                                        OS
                                       OR
                       Rapidly Perfused
                                           CA
                                                           RUR
                             Metabolites
    Fig. 1. Scheme of physiologically based pharmacokinetic model (Pb-Pk) used to simulate
lactational transfer of tetrachloroethylene (PCE) in nursing rats.
    Abbreviations: CI = concentration in inhaled air (mg/L); QP = alveolar ventilation rate
adjusted for body weight (L/h); CX = concentration in  exhaled air (mg/L); QC = cardiac
output adjusted  for body weight (L/h); CVF  =  venous  concentration leaving the fat tissue
(mg/L); QF = blood flow to fat (L/h); CVS = venous concentration leaving the slowly perfused
tissues (mg/L); QS = blood flow to slowly perfused tissues (L/h); CV  = concentration in mixed
venous blood (mg/L); CA  =  concentration  in arterial blood  (mg/L);  CVR  = venous
concentration leaving the rapidly perfused tissues (mg/L); QR = blood flow to rapidly perfused
tissues (L/h); CVL = venous concentration leaving the liver tissue (mg/L); QL = blood flow
to liver (L/h); QPP = alveolar ventilation rate in pups adjusted for body weight (L/h); CXP =
concentration  in air exhaled  by pups (mg/L); CVMT = venous concentration leaving the
mammary glands tissue (mg/L); CMAT = concentration in milk (mg/L); QCP = cardiac output
in pups adjusted for body weight (L/h); CVP = concentration in venous blood in pups (mg/L);
CAP = concentration in arterial blood in pups (mg/L); RPUP = elimination rate for PCE from
milk to pups (mg/h); KFC = first order metabolism rate constant (1/h/kg); VMAX = pseudo-
maximal velocity of PCE metabolism (mg/h); KM = apparent Michaelis-Menten constant for
PCE metabolism (mg/L); AM = amount of PCE metabolized (mg); RMR = the rate of
gastrointestinal tract loading with PCE in pups (mg/h); RAP = the rate of gastrointestinal
absorption of PCE in pups (mg/h).

-------
TABLET.   KINETIC  CONSTANTS  AND
PARAMETERS USED IN PB-Pk MODELING OF
TRANSFER OF TETRACHLOROETHYLENE IN  RATS
HUMANS
Tissue Volumes
  Maternal
    Liver
    Fat
    Mammary
  Perinatal
    Pup (Infant)  Tissue

  Maternal
    Slowly Perfused
    Rapidly Perfused
    Milk Volume

 Flow  Rates
  Maternal
    Alveolar Ventilation
    Cardiac Output
  Perinatal
    Alveolar Vent. Pup (Inf
    Cardiac Output Pup (Inf

  Maternal
    Liver
    Fat
                           '[Fraction of Body WeigntT
                                VLC  =0.04
                                VFC  =0.1
                                VMATC= 0.044

                                VTCP =0.9
                                              0.04
                                              0.2
                                              0.05

                                              0.9
                                 VS   =  0.79*BW-VF-VMAT
                                 VR   =  0.12*BW-VL
                                 VMILK=  0.00233      0.03542
                            [L/h/kg]

                                 QPC
                                 QCC
                                = 14.5
                                = 14.3
             19.7
             18.0
                       )    QPCP =30.0
                       )    QCCP =22.0
                       [Fraction of Cardiac Output]
                                                    25.2
                                                    22. .0
                                 QLC
                                 QFC
                                = 0.25
                                = 0.07
Partition Coefficients
  Maternal
    Blood/air
    Liver/blood
    Fat/blood
    Slowly Perf./blood
    Richly Perf./blood
    Milk/blood
  Perinatal
    Blood/air Pup (Inf.)
    Other Tiss./bld.Pup (Inf.)

 Metabolism
                             [Ratio of Solubility]
                                  PB
                                  PL
                                  PF
                                  PS
                                  PR
=33.
= 1.
=42.
= 0
= 1
                                    5
                                    9
                                    35
                                    93
                                    67
                                  PMILK=12.0
                                  PPB
                                  PPT
                        [mg/t,
Apparent Michaelis-Menten    KM
 *                      [mg/kg/h]
Pseudo Maximal Velocity    ..I™*
                        [1/h/kg]
First Ord. Metab.  Rate      KFC
 =24
 = 4
                                    .3
                                    .54
                                       - 0.32

                                         °'°3
                                       = 1.2
              0.25
              0.05
 19.8
  6.83
159.03
  7
  6
  2
                                                       77
                                                       83
                                                       8
  8.
  6
                                                       0
                                                       596
               0.32

               0.151

               1.2

-------
the mammary glands.  Concentration, CMAT,  equals to AMAT/VMILK where  VMILK
represents volume of milk. It was assumed that the milk compartment is in intimate contact with
the arterial blood perfusing the mammary tissue, and that PCE rapidly equilibrates with the milk.
    The rate of change in  the amount of PCE in the pup's gastrointestinal tract (AGIT) is
described as a difference between the rate of ingesting of PCE with mother's milk (RPUP) and
the rate of absorption from the gastrointestinal tract, RAP (equal to MR*KAP; where MR is the
amount remaining in the gastrointestinal tract of pup, KAP  is absorption constant for pup,
determinated to be equal to 0.5 1/hr):

             dAGIT/dt =  (RPUP-RAP)

The concentration of PCE in pup's gastrointestinal tract (CGIT) was calculated as MR/GIW,
where GIW represents weight of gastrointestinal tract of pup, adjusted for the pup's weight.

3.3. Closed Chamber Exposure
    The closed chamber gas uptake data (Fig.  2)  were  used to estimate  and optimize the
metabolism constants (VMAXC = 0.03 mg/kg/h; and KM = 0.32 mg/L). These values suggest
a very slow metabolism rate of PCE by lactating rats.  Kinetic constants and physiological
parameters  are listed in Table 1.
    Computer simulations of gas uptake exposure to the initial air concentration of 670 ppm of
PCE for 6 h were conducted and predictions of the PB-Pk model are shown in Figure 3.  The
predictions  of blood  and milk levels were compared to the results of measurement of respective
concentrations at the end of  exposure (CV = 6.14±0.29 mg/L; CMAT  = 89.18±12  mg/L; n
= 3).

3.4. Open Chamber Exposure
    Further validation  of the PB-Pk model was completed using lactating rats exposed for 2
hours to constant concentrations  of PCE (ranging from 20jf2  to  1000.+47  ppm).  The
dependence of PCE  concentration in  milk versus concentration in air (CI) is shown in Fig. 4.
The data for blood and milk collected from rats  exposed to these different concentrations of PCE
were compared to the simulated levels by PB-Pk model. Validation of the  dose-dependent model
predictions  is shown in Fig. 5. Similarly, the data collected  from lactating rats exposed for
different time to constant concentration of 600 ppm of PCE were compared to the levels in blood
and milk simulated  by  PB-Pk model. Validation of the time-dependent model predictions is
shown in Fig. 6. In  both cases, the  dose- and time-dependent courses of PCE concentrations
were in agreement with those predicted by the PB-Pk model.

3.5. Exposure of Pups via Mother's Milk
    Another  group of lactating  rats was  exposed for  2 hours  using  the same exposure
concentration of 600 ppm of PCE. The dams were next returned  to their nursing pups and
concentrations of PCE were followed for the next 24 hours in blood and milk. Validations of
the time-dependent model predictions are shown  in Fig. 7a and 7b,  respectively. To achieve a
clearance of PCE from the systemic  circulation at 26 hours, as indicated by the experimental
data presented in Fig. 7, a small first order term KFC was introduced (=1.2 1/hr/kg; see Table
I).
    Concentrations  of PCE were measured also in 10 to 11 day old pups for up to 24 hours

-------
                                         Rat Lactation—Inhalation Gas Uptake
   Rat Lactation—inhalation Gas Uptake

   10*
                                         300
                                                      1*4
                                                        Tim* (hours)
    Fig 2 Results of gas uptake measurement (small rectangles) and
(solid ^) during exposure of lactating rats to eight different initial co
    aS 1000 ppi of tetracMoroethylene (PCE) for 6 hours (n=3 for each data pornt).
    Fie 3  Validation of PB-Pk model predictions (solid lines) of

ZStt~!2^KXtt3Z*-*'-. -
standard deviation (n=3).

-------
                                                       Rat Lactation—Inhalation
 Rat Lactation—Inhalation, Dose-Dependent
    MO
 ,£ 800
                         OJ  .   OJJ
            C (10s) PCE Concentration (ppm)
                                                        0.4    &*    1£    I*
                                                                 Time (hour*)
Z.4   2.8
                                                                 Time (hours)
     Fig. 4. Relationship between concentration of tetrachloroethylene (PCE), measured in milk
shortly after exposure, and the time weighted average PCE concentration in inhaled air for rats
exposed for 2 hours to 20 ppm to 1000 ppm of PCE (n=5 for each PCE air concentration).

     Fig. 5. Validation of PB-Pk model predictions (solid lines) of tetrachloroethylene (PCE)
concentrations in blood (A) and milk (B) of lactating rats (small rectangles) exposed to different
constant concentrations, ranging from 20 ppm to  1000 ppm of PCE for 2 hours (n=5 for each
PCE air concentration).
                                           9-

-------
        Rat Lactation—Inhalation
   10*
I  w'
5j
             I-CY ;

A  "
                                           Ul
                                                    Rat Lactation—Inhalation, Dam
                 1     i     4     §     «
                   Time (hours)
                                                                 Tlma (hours)
                                               800
                                                               to     if      »
                                                                  Time (hours)
                                                                                          30
        Fig  6 Validation of PB-Pk model predictions (solid lines).of tetrachloroethylene (PCE)
    concentrations in blood (A) and milk (B) of lactating rats (small rectangles) exposed to a constant
    concentration of 600 ppm of PCE for 2 hours to 5 hours (n=5 for each exposure tome).


         Fig  7   Validation  of PB-Pk model  predictions  (solid  lines)  of  time-dependent
    tetrachloroethylene (PCE) concentrations in blood (A) and milk (B) of lactating rats (small
    rectangles) exposed to 600 ppm of PCE for 2 hours (n=2 for each recovery tome).
                                                10

-------
 after the exposed dams were returned to feed them. Model validation of the time-dependent
 predictions for pups is shown in Fig. 8 and 9. The nursing pup whole body burdening of PCE
 was slightly underpredicted (Fig. 8a). On the other hand, PCE concentrations in venous blood
 of pup were slightly overpredicted for times longer than 6 hours (Fig. 8b). From comparison
 of the peak PCE concentration in a whole pup, including gastrointestinal tract (Fig. 8a), with
 the peak concentration, of PCE in  venous blood (Fig.  8b)  it seems that the  apex  in  blood
 appeared several hours later than that in a whole pup. Even later than in blood, the apex of PCE
 concentration appeared in solid pup tissues, other than gastrointestinal tract (Fig. 9a). On  the
 other hand, peak loading of the gastrointestinal tract (Fig. 9b) with PCE ingested by pups with
 milk, appeared  about five hours earlier than the peak in solid tissues. The loading  of pup's
 gastrointestinal tract and other solid  tissues was slightly underpredicted by the model (Fig. 9).

 4.  DISCUSSION

 4.1. PB-Pk Models in Lactational  Transfer of Chemicals
     The implementation of physiologically-based pharmacokinetic models by Shelley etal. (14,
 17) represented significant progress in estimating the infant's exposure to chemicals transferred
 with breast milk and assessment of the overall risk to the infant. Their approach involved a
 physiologically-based mathematical simulation capable of modeling, for instance, the distribution
 of  volatile organic solvents  from   mother's  breathing  zone to the nursing  infant.  Such
 physiologically-based pharmacokinetic models of lactational transfer of chemicals may be scaled
 up or down, according to the body weight, and can be validated using laboratory animals, such
 as lactating rats (15).
     The same approach was used in the present study. However, the PCE distribution in dam
 was better  described by five compartments rather than three, as in the general PB-Pk model
 presented by Shelley et al. (14, 17). On the other hand,  the pup was tentatively described by
 one, and finally by two compartments  only, without incorporating metabolism. This was in
 contrast to  the PB-Pk model for trichloroethylene by Fisher et al.  (15). Elimination of PCE in
 the pup was assumed to occur by exhalation.  The physiological parameters, partition coefficients
 and metabolism parameters estimated or determined by experiments are shown in Table  I. Using
 these parameters,  the PB-Pk  model   described fairly  accurately both blood and  milk
 concentrations of PCE in lactating rats exposed to different concentrations of this chemical for
 different periods of time (Fig. 3, 5,  6).
     On the other hand, differences between the kinetics of loading pup's gastrointestinal tract
 and blood or solid tissues required a  separation of the gastrointestinal tract as a separate, initial
 compartment which is loaded with PCE about 5 hour prior to the venous blood and solid tissues
 (Fig. 8, 9). Using these assumptions, the Pb-Pk model predicted fairly accurately and reliably
 the distribution of PCE inhaled by dams and passed onto  their nursing pups via breast milk.

4.2. Computer Simulation of Repetitive Exposures in Rats
    Assuming 2 hours exposure to 600 ppm of PCE, five times per week, a prognosis  was run
to predict the time-course of PCE concentration in rat's milk for one month (Fig. lOa). With* the
same model settings, the accumulated dose of PCE received by 8 pups was estimated to reach
as much as  85 mg within one month  (Fig. lOb).  Under these exposure conditions it seems very
likely that the concentration of PCE in pup's blood may reach 1 mg/L, which is a concentration
referred to by the American Conference of Governmental Industrial Hygienists (ACGIH) as the
                                           11

-------
Rat Lactation—Inhalation, Pup Average
                                                 Rat Lactation—Inhalation, Pup Average
              to     ts     20
                 Time (hours)
0.0
               10     15
                  Time (hour*)
18      15      »
   Time (houra)
      Fig.  8.  Validation  of PB-Pk  model  predictions (solid  lines)  of time-dependent
 tetrachloroethylene (PCE) concentrations in tissues of whole pup (A) and venous blood (B) of
 pups fed by the dams exposed to 600 ppm of PCE for 2 hours. The small rectangle and vertical
 bars show mean ± standard deviation (n=3 or 6 for each recovery time).

      Fig.  9.  Validation  of PB-Pk  model  predictions (solid  lines)  of time-dependent
 tetrachloroethylene (PCE) concentrations in solid tissues (A) and gastrointestinal tract (B) of pups
 fed by the dams exposed to 600 ppm of PCE for 2 hours. The small rectangle and vertical bars
 show mean ±. standard deviation (n=3 or 6 for each recovery time).
                                            12

-------
 index of biological exposure (BEI) to the threshold limit concentration (TLV-TWA = 50 ppm)
 for PCE inhaled by adult human subjects (18).

 4.3.  Computer Simulations and Prognosis of PCE Distribution in Humans
      The attempt  was made to scale-up  the PB-Pk model and to test its predictions  versus
 available data for humans. Initially, a set of physiological parameters and kinetic constants,
 pertinent to PCE in humans, was adopted from Ward et al. (10). The values describing  human
 milk and mammary glands compartment  were calculated from data published for "Reference
 Man" (19), and finally the constants were optimized over the literature data describing inhalation
 exposures  of human subjects to PCE (7-9, 20, 21). The final set of parameters and constants
 is listed in Table I.
      Using these parameters with the milk compartment turned off, and the exposure scenario
 described by ACGIH  (18), the computer simulations of PCE concentrations in blood (Fig. lla)
 and  exhaled air (Fig.  lib)  were  run and compared to  BEI values. The model slightly
 underpredicted both blood and exhaled air PCE concentrations for human subjects, prior to the
 last shift of the workweek (Fig. 11). Much better fit of the computer-simulated time-course was
 obtained with the  data reported by Fernandez et al.  (20) for exhaled air of human subjects
 exposed to 100 ppm of PCE for 1  hour (Fig. 12a) and 8 hours (Fig. 12 b). Similarly, the model
 predicted accurately PCE concentrations in exhaled air of human subjects exposed to 194 ppm
 of PCE for 90 min and 3 hours  (Fig. 13), as reported by Stewart et al. (7). Fig. 14  shows
 computer  simulation  of the  rate of  PCE exhalation (RAX) run versus data reported  by.
 Bolanowska et al.  (21) for two human subjects, a slim man and an obese woman (Bolanowska,
 personal communication). The model predictions of PCE exhaled breath clearance rates (RAX)
 in the two subjects were in general agreement with modest overprediction of experimental data
 after the first measured time point (Fig. 14).

 4.4. Computer Simulations and Prognosis of PCE Distribution in Mother and Her Nursing
 Infant
     An attempt was made to simulate the only documented case of the lactational transfer of
 PCE from mother  to infant, described by Bagnell et al.  (5). Although the PCE concentration in
 inhaled air was not measured the reported incidents of dizziness after exposure of mother to PCE
 without symptoms  of general anesthesia (5) suggested the PCE air concentration within the range
 of  several  hundreds ppm. The best approximation of the computer simulated values to PCE
•concentrations determined in  blood and breast milk  (5)  was achieved when the exposure
 concentration in inhaled air was  assumed to be 600 ppm (Fig. 15a, b). This concentration,
 exceeding more that ten times the air TLV-TWA level recommended by ACGIH (18) for PCE,
 could result in the infant blood concentration of not more than 0.035 mg/L within one  month
 of exposure to PCE via mother's milk. This concentration is more  than one order of magnitude
 lower than the no-effect threshold assumed for adults by ACGIH (18).
     From these several prognoses of PCE distribution and its kinetic behavior in exhaled air,
 blood and  milk of exposed human subjects and especially from the comparison of computer
 simulations with the available human data from literature, it is concluded that the PB-Pk  model
 described fairly accurately  the concentrations of PCE in both lactating rats and humans. It
 seems that the validated PB-Pk model may be used, consequently, to predict the absorbed doses
 of PCE by nursing infants from the concentration in mother's breathing zone. Although this
 approach will require  monitoring of concentrations of PCE and other volatile chemicals  in the
                                            13

-------
     Rat Lactation—Inhalation, Pup Average
   (00
-7  500
   400
   000
8
UJ
   100
   100
       CMAT
           100    200    300    400    500    600
                      Time (hour*)
B
                       Time (hours)
                                             TOO
                                                  B
                        300    400    500    *00    TOO
                                                       30
I
Is
1
I
S   15
e
o
                                                       10
                                                    3
                                                    Ul
                                                    o
                                                    a.
                                                                  ACGIH. TLV-BE1
                                                     25     SO
                      75    100    125    150    175
                     Time (hours)
                                                                     /
                                                                 ....A
                                                                                     .CXPPM
                                                               25
                                                           15     7S    100    125
                                                                 Time (hours)
   Fie 10 Simulatedprognosisoftime^ependentconcentrationsoftetrachloroethylene(PCE)
 raSVuk (^Sumulated doses of tetrachloroethylene (PCE) received by 8 pups wi*
iS from Ae dams (B). The computer simulation was run assuming 2 hours ^sure of dams
         of PCE, five times per week (Monday through Friday, beguung on Fnday), during
       to
       1 month.
            Fig. 11. Simulatedprognosis
                     frreflmeipet week for 8 hours, Monday through Friday, beguung on Monday).
                    i^ lowSgical exposure mdie« (BED in blood (A) and exhaled a* (B). Data
        according to (18).
                                                 14

-------
Fernandez et al. 8 Hours Exposure to 100 ppm of PCE Stewart et al. Exposure: ••-90 Minutes; •—3 Hours
 A  "I	'
     30
  Ul.
  o
  a.
5   eo
I
I
S   40
                                                     2
                                                     i
                                                        20
Ul
O
a.
                                                                                  • • = CXPPM
                       10       12
                       Time (hours)
                   3      12      16      20     24
                     Time (hours)
                                               Bolanowska et al. Subject:  •—Obese Woman; •—Slim Man
                                                       100
 Fernandez et al. 1 Hour Exposure to 100 ppm of PCE
B 4U
Exhaled A
is
c
g
|
§ 10
UJ


^






D 1

; B= CXPPM


• i :
1 	 ' 	 | 	 '"] 	

234!
                                                                         12       16
                                                                         Time (hours)
                        Time (hours)
             Fig. 12. Simulated prognosis of time-dependent concentrations of tetrachloroethylene (PCE)
         in exhaled  air of human subjects exposed to 100 ppm of PCE for 1 hour (A) and 8 hours (B),
         according  to the scenario reported by Fernandez  et al. (20). Small rectangles show data in
         exhaled air according to  (20).

             Fig. 13. Simulated prognosis of time-dependent concentrations of tetrachloroethylene (PCE)
         in exhaled  'air of human subjects exposed to 194 ppm  of PCE for 1.5 and 3 hours, according
         to the scenario  reported by Stewart et al. (7). Small rectangles show  data in exhaled air
         according to (7).

             Fig.  14. Simulated prognosis of time-dependent rates of   tetrachloroethylene  (PCE)
         exhalation  by obese woman and slim man, exposed to 55 ppm of PCE for 6 hours, according
         to the scenario reported by Bolanowska et al.  (21). Small rectangles show data in exhaled air
         according to (2-1).                                                    --^~»-
                                                     15

-------
Bagnell et al. Nursing Mother Exposure to 600 ppm of PCE
                                              "~
        10
                                                                       200    300    «0
                                                                             Time (hours)
                            12    18
                           Time (hours)
      'according to (21).
       1 month, according to the scenario reported by Bagnell et al. (21).
                                                   16

-------
 air, it would aid the future attempts of risk assessment for infants.
 ACKNOWLEDGEMENTS
     This work was performed under Department of the Air Force Contract #F33615-90-C-0532.

 REFERENCES

 1.  M.V.Cone, M.F.Baldauf, D.M.Opresko, and  M.S.Uziel, "Chemicals identified in human
   breast milk, a literature search" (EPA 560/5-83-009 Report. US. Dept. of Commerce Natl
   Technical Inform. Service, Washington, DC, 1983).
 2. D.Giroux, G.Lapointe, and M.Baril, "Toxicological index and the presence in the workplace
   of chemical hazards for workers who breast-feed infants," Am. Ind. Hyg: Assoc.  J. 53
   471-474.
 3. A.A.Jensen, S.A.Slorach, "Chemical Contaminants in Human Milk" (CRC Press Inc  Boca
   Raton, FL, 1991).
 4.  Committee on Drugs, American Academy of Pediatrics,  "Transfer of drugs and other
   chemicals into human milk," Pediatrics 84, 924-936 (1989).
 5. P.C.Bagnell, and H. A.Ellenberger, "Obstructive jaundice due to a chlorinated hydrocarbon
   in breast milk," Can. Med. Assoc. J. 117, 1047-1048 (1977).
 6. R.D.Stewart, "Acute tetrachloroethylene intoxication," J. Amer. Med. Ass. 208  1490-1492
   (1969).
 7, R.D.Stewart, E.D.Baretta, H.C.Dodd, and T.Torkelson, "Experimental human exposure to
   tetrachloroethylene,"  Arch. Environ. Health 20, 224-229 (1970).
 8.  R.D.Stewart, D.S.Erley, A.W.Schaffer, and H.H.Gay,  "Accidental vapor exposure to
   anesthetic concentrations of a solvent containing tetrachloroethylene," Ind. Med Surg  30
   327-330(1961).                                                                  '
 9. R.D.Stewart, H.Gay, D.Erley, C.Hake, and A.Schaffer,  "Human exposure to
   tetrachloroethylene vapor," Arch. Environ. Health 2, 516-522 (1961).
 10. R.C.Ward, C.C.Travis, D.M.Hetrick, M.E.Andersen,.andM.L.Gargas, "Pharmacokinetics
   of tetrachloroethylene," Tox. Appl. Pharmacol. 93, 108-117 (1988).
 11. M.L.Gargas, M.E.Andersen, and H.J.Clewell,m, "A physiologically based simulation
   approach for determining metabolic constants from gas uptake data," Toxicol Appl
   Pharmacol. 86, 341-352 (1986).
 12. M.L.Gargas, RJ.Burgess, D.E.Voisard, G.H.Cason, and M.E.Andersen, "Partition
   coefficients of low-molecular weight volatile chemicals in various liquids and tissues,"
   Toxicol. Appl. Pharmacol. 98, 87-99 (1989).
 13. J.C.Ramsey, and M.E.Andersen, "A physiologically based description of the inhalation
   pharmacokinetics of styrene in rats and humans. Toxicol. Appl. Pharmacol. 73. 159-175
   (1984).
14. M.L.Shelley, M.E.Andersen, and J.W.Fisher,  "An inhalation distribution model for the
   lactating mother and nursing child," Toxicol.  Lett. 43, 23-29 (1988).
15. J.W.Fisher, T.A.Whittaker, D.H.Taylor, H.J.Ciewell,m, and M.E.Andersen,
   "Physiologically based pharmacokinetic modeling  of the lactating rat and nursing pup: A
   multiroute exposure model for trichloroethylene and its metabolite, trichloroacetic  acid.
   Toxicol. Appl. Pharmacol. 102, 497-513 (1990).
16. J.Z.Byczkowski, E.R.Kinkead, R.J.Greene, L.A.Bankston, and J.W.Fisher,
                                        17

-------
"Physiologically-based modeling of the lactational transfer of tetrachloroethylene,"

 .
(1972).
                                "inhalation
 experimental conditions," Medycyna Pracy 23,
                                            1-8

-------
  Biographical Sketch:  MAJ(P) Daniel J. Caldwefl, Ph.D., U.S. Army
MAJ Caldwell received his undergraduate degree in Environmental Health from East
Tennessee State University in 1976.  He also holds a M.S. in business administration
from Boston University Metropolitan College (1981) and a M.H.S. in Occupational
Safety and Health from the Johns Hopkins University School of Hygiene and Public
Health (1984).  His Ph.D. in Toxicology was received from the University of Pittsburgh
in 1991. MAJ Caldwell is certified as a Diplomate of the American Board of Industrial
Hygiene.

Since May of 1991 he has served as Chief of the Toxicology and Exposure
Assessment Section, Occupational Health Research Detachment, of the U.S. Army
Biomedical Research and Development Laboratory. He is responsible for health
hazard and exposure assessment, development of a combustion toxicology database,
and formulation of permissible'exposure standards for military unique compounds or
substances for which there are no federal standards.  MAJ Caldwell has also been
assigned to the Army Environmental Hygiene Agency and the Ft. Meade Medical
Department Activity, where he was responsible for health hazard assessments and
environmental and occupational health programs at numerous Army installations. He
has authored/co-authored .several publications, a chapter in  a medical text book, and
two Army Technical Bulletins. MAJ Caldwell is an  Adjunct Assistant Professor of  .
Engineering  and Environmental Management at the U.S. Air  Force Institute of
Technology.

-------
  Issues in the Development of a Risk Assessment for Fire Hazards
                       MAJ(P) Daniel J. Caldwell, Ph.D.
      In  general terms,  the risk assessment process determines  information  on the
nature and extent of a hazard, as differentiated from the risk management process which
encompasses judgements of acceptability and determination of degree to which the risks
should/can  be controlled [1].   Risk assessment is  primarily scientific,  while  risk
management takes into consideration other factors such as cost, technicai feasibility, and
timing.
      Numerous hazards to humans result from exposure to fire conditions and the
products of combustion. Predominant among these are effects from heat and flames,
visual obscuration due to the density of smoke or to eye irritation, narcosis from inhalation
of asphyxiants, and irritation of the upper and/or lower respiratory tracts [2].  These
 effects often occur simultaneously in a fire and contribute to physical incapacitation, loss
 of motor coordination, faulty judgement, disorientation, restricted vision, and panic. The
 resulting delay or prevention of escape may lead to subsequent injury or death from
 further inhalation of smoke.
       Assessing the effects of exposure of humans to smoke is extremely difficult since
 smoke is a continuously changing mixture of airborne solid and liquid particles and gases
 which is evolved when a material undergoes pyrolysis or flaming combustion.  The toxicrty
 of the smoke produced is therefore time-dependent [3,4]. This dictates that a dynamic
  (i.e., flow-through) animal exposure system be employed to properly evaluate the toxicity

-------
of the smoke [5].  To fully assess fire hazard, one must evaluate the inherent material
properties that influence flannmability and flame spread and integrate this information with
toxicity data obtained from a combustion test method which replicates likely real-world
exposures [6,7]. The two variables affecting the toxicity of combustion products are the
yield of toxicants and the tirne-to-effect [6].  Historically, death has been the endpoint of
choice for evaluation of the toxic potency of combustion products [4,7].  However, this
does not account for the time to effect.
      To  minimize risk from fire hazards, emphasis must be placed on SURVIVABLE
EXPOSURES rather than lethality. The survival time is measurable and dependent upon
two variables: potency of each toxicant evolved as well as their rapidity of action. Burning
conditions influence both the evolution of toxicants and behavior of a material under fire
conditions. Thus, the inherent chemical/physical properties which determine flammability
also effect the toxic hazard of a material. Because of this,  toxicity of smoke can not be
predicted  a priori  by simple  additive models based on the fractional effective dose
approach  (e.g., the N-gas model, which while suitable for multiple .gas exposures it fails
to account for the effects due to smoke exposure).
      Fortunately, however, a methodology exists  which  has been proven to predict
toxicity due to smoke exposure [5]. This method is an evolutionary development in the
field of combustion toxicology which allows, for the first time, an evaluation of materials
under well defined burning conditions over a wide range of heat flux and ventilation levels.
The toxic  hazard of the smoke produced was influenced by the burning conditions to
which the materials were exposed.

-------
    •  A  Potential Smoke  Hazard  index was  developed  to  integrate the  material
performance characteristics, which determine ease of ignition and flame spread, with the
toxic hazard (i.e., toxic potency, time to toxic effect, and rate of generation of toxicants).
Not only is the toxicity of the smoke described in terms of toxic potency  (i.e., LC50), but
the time, to toxic effect and fundamental  material  properties that influence the burning
behavior are also determined [6].
       Furthermore, the influence of a given set of burning conditions on the time to toxic
effect, e.g., death or incapacitation, has never been evaluated in detail [2].  Hazardous
concentrations of smoke develop in  a fire prior  to "flashover", thus it  is important to
determine at what time an untenable  condition  occurs.  The time to manifestation of a
toxic effect is measurable and dependent upon  two variables: 1) the rate of evolution of
toxicants from the burning material, and 2) the potency and rapidity of action of each
 toxicant  generated.   The burning conditions  directly  influence the rate  of toxicant
 evolution.
       Three fundamental variables are needed  to describe the burning conditions of real
 fires. These are the imposed heat flux, or irradiance, the ventilation, and the mass loss
 rate [5]. Previous small-scale methods to evaluate toxicity of smoke are restricted to one
 set of conditions in  each apparatus used. In order to evaluate the effects of burning
 conditions on smoke toxicity, a small-scale flaming combustion apparatus was developed.
 The UPITTII apparatus, which has been described previously in great detail, is unique in
 that it permits control of both heat flux and ventilation, and measurement of the mass loss
  rate of the burning specimen [5],

-------
       As discussed above, the time to effect plays a major part in determining the toxic
 hazard of smoke. Although the toxic potency of a material changed very little over the
 range of burning conditions investigated, the time to effect was greatly influenced by the
 burning conditions.  The relationship between the median time to death and the smoke
 concentration was investigated to obtain an estimate of the time necessary to produce
 an effect starting at the beginning of heat flux to the specimens.
       Material performance data have not been considered, except the mass loss rate
 [4].  It is clear that material performance should be included in the risk analysis since the
 stability of some materials precludes ignition at lower imposed heat flux levels; yet this fact
 is usually neglected, and taking toxic potency alone is not appropriate when addressing
 the total hazard from fires. The hazard analysis must include material performance as.
 well  [5,6].  The time to ignition (Tign) for a material at a variety of heat flux levels yields
 data from which the critical flux (q"cr), which is the flux below which no ignition or pyrolysis
 will occur, and thermal inertia (Tl) of a material, can be determined [5].  These variables
 can then be taken into account to calculate an index of potential smoke hazard  (PSH) [6].
       The PSH index incorporates the variables, q"cr and Tl, which determine "ease of
 ignition" of a material and rate of fire development or flame spread, the toxic potency and
time to effect, and the rate of generation of toxicants from the mass loss rate (m).
The PSH index can be modified to permit calculation of an index based on survivability
data obtained from studies using such alternative endpoints, while retaining the material
performance characteristics determined with this method [8].  Such an index would be
a valuable tool to .assist in materials selection decisions.

-------
f NSAcademy of Sciences (1983).  Risk Assessment ir .the Tedera, Government:
Managing the Process.  National Academy Press, Washington, DC. pp. 1^1.
2. National Research Councii, Committee on Fire ^c^(^6.e and Smoke:
Understanding the hazards.  National Academy Press, Washington, DC.

3  International Organization for Standardization (1992). Technical Report TR-9122, Part
4: The Fire Model, ISO/TC92.

4  International Organization for Standardization (1992). Technical Report TR-91 22, Part
5: Prediction of toxic effects of fire effluents, ISO/TC92.
(1990).

iomcs^^^
UPfc U Fiamfng Combustion/Toxicity of Smoke Apparatus. J. of F.re Sciences 9, 470-518,
(1991).
7  Purser D A 1992. "The harmonization of toxic potency data for materials obtained

OR September 1992. Interscience Commumications, LTD, Lonaon.
                        "Application  of the UPitt
                             39th
n
decisn:
                                                method to materials selection
                                                        Research^fereDc
 14-18 September. 1992. in press
 	The views opinions, and/or findings contained in this report are those of the
 authored should not be construed as official Department of the Army poabon, pohcy,
 or decisTon, unless so designated by other official documentor,.

-------
           BIOGRAPHICAL SKETCH: HARVEY J. CLEWELL, III

Harvey J.  Clewell,  III is  currently Senior Project  Manager and
Director of the Health Assessment Group for the  K.S. Crump Division
of ICF Clement International.   in  this  capacity,  he conducts and
directs research to assess the health risks associated with toxic
chemicals, as we'll as to advance the state of the art of chemical
risk assessment.   He  recently retired from the  Air Force  as  a
Lieutenant  Colonel;  during  his Air Force  career he  performed
research  in  environmental  modeling and  in physiologically based
pharmacokinetic  modeling   for  application   to   chemical  risk
assessment.    His  positions  in the  Air  Force  included  Deputy
Director of the Toxic Hazards Division  (the Air Force's toxicology
unit) and Director of Hazardous Materials Safety for Aeronautical
Systems Division.  He  also  served  as Consultant to the Air Force
Surgeon General for chemical risk assessment.

-------
INCORPORATION OF PHARMACOKINETICS IN NON-CARCINOGENIC RISK ASSESSMENT:
                   EXAMPLE WITH CHLOROPENTAFLUOROBENZENE

                         Harvey J. Clewell IIP and Bruce M. Jarnotb

                  a K.S. Crump Division, ICF International, Ruston, Louisiana
     ' " Toxicology Division, Armstrong Laboratory, Wright-Patterson Air Force Base, Ohio

                                        ABSTRACT

       Non-carcinogenic risk assessment traditionally relies on applied dose measures, such as.
 concentration in inhaled air or in drinking water.  Safety factors are then incorporated to address the
 uncertainties associated with  extrapolating across species, dose levels, and routes of exposure, as well
 as to account for the potential impact of variability of human response.  A risk assessment for
 chloropentafluorobenzene (CPFB) was performed in which a physiologically based pharmacokinetic
 (PBPK)  model was employed to calculate an internal measure of effective tissue dose appropriate to
 each toxic endpoint.  The model accurately describes the kinetics of CPFB in both rodents and
 primates.  The model calculations of internal dose at the no-effect and low-effect levels in animals
 were compared with those calculated for potential human exposure scenarios.  These calculations were
 then used in place of the default safety factors to determine safe human exposure conditions.
 Estimates of the impact of model parameter uncertainty and human pharmacokinetic variability, as
 estimated by a Monte Carlo technique, were also incorporated into the assessment.  The approach
 used for CPFB is recommended as a general methodology for non-carcinogenic risk assessment
 whenever the necessary pharmacokinetic data can be obtained.

                                       INTRODUCTION

         For a number of years the U.S. Air Force has been performing research to develop safe
  intake simulants for chemical warfare agents, in order to provide accurate and quantitative real-time
  assessment of troop proficiency and gear efficacy during chemical warfare field exercises.
  Chloropentafluorobenzene (CPFB) was identified and  evaluated as a candidate inhalation simulant,
  and was determined to possess desirable physicochemical and toxicological properties. These include
  rapid uptake, low metabolism and toxicity, rapid and  predictable clearance, real-time detectability by
  existing portable "breathalyzer" technology, gas mask breakthrough similar to the actual agents, and

-------
commercial availability. Before using CPFB in human trials, it was important to determine safe
exposure conditions, taking into consideration the exposure levels at which toxicity was observed in
animal studies.

        Because of the need to balance protection of personnel during training with the ability to
provide effective training for a dangerous wartime scenario, an accurate (as opposed to simply safe-
sided) estimate of acceptable human exposure was needed.' The usual practice for non-carcinogenic
risk assessment (Barnes and Dourson,  1988) uses measures of applied dose to relate to toxicity.
Safety factors are then applied to account for  uncertainty regarding the relationship between applied
dos and effective target tissue dose across routes of exposure and species, as well as for variability in
the human population.  A more scientifically based approach would be to use a measure of tissue dose
directly and to use known principles of pharmacokinetics to relate different exposure scenarios.  For
this purpose a physiologically based pharmacokinetic (PBPK) model was developed which could be
used to  perform the route-to-route and cross-species extrapolations necessary to develop a human risk
estimate.  The model  was also used  in a  Monte Carlo analysis to estimate the uncertainty and
variability associated with the risk estimate.

                                PBPK  MODEL .DEVELOPMENT

Structure

        The structure  of the model is shown in Figure  1, and the assumptions underlying the
mathematical description follow those of Ramsey and Andersen (1984) with the following exceptions:

        1. The model of Ramsey and Andersen included only saturable metabolism. An additional
        pathway of metabolism has been added in this model which is linear in concentration. Thus
        the equation for the rate of change of amount of CPF.B in the liver contains an additional term:
                - KF * CL * VL / PL
(where the parameters are defined in Table 1)
       2. A GI tract compartment has been added. Oral absorption takes place into this compartment
       by a first order process: KA * AST (where AST represents the amount of CPFB remaining in
       the stomach). The liver receives the blood flow from this compartment (QG) as well as its
       own arterial supply (QL).  Thus the equations for the rate of change in the amount of CPFB in
       the stomach (RAST), GI tract (RAG), and the liver (RAL) are:

-------
             RAST = - KA * AST
             RAG = QG * (CA - CG / PG) 4- KA * AST
             RAL = QG * (CG / PG -  CL / PL) + QL * (CA - CL / PL)
                    - VMAX * CL / PL / (KM + CL / PL) - KF * CL * VL / PL

      3.  A bone marrow compartment has been added.  The form of the equation for the rate of
      change in the amount of CPFB in the bone marrow (RAM) is identical to that of the other
      basic tissues in Ramsey and Andersen (e.g., fat, slow, rapid):
             RAM = QM * (CA - CM / PM)

      4. In order to better simulate the measurements of exhaled breath in anesthetized monkeys,
      the description of gas exchange between the lung and the blood was modified to explicitly
      model an alveolar space in which inhaled air at concentration CI and air in equilibrium with
      the blood were mixed.  In place  of the steady-state assumption used in Ramsey and Andersen,
      the following rate equation for the  amount of CPFB in the blood (ABL) was integrated along
      with the equations for the tissue  compartments:
              RABL = QP * (CALV -  CX) + QC * (CV - CA)
      where:
              CALV = ALVS * ABL / (VBL*PB) + (1. - ALVS) * CI
      The measured exhaled air concentration in parts per million (CXPPM) was then described by
       the equation:
              CXPPM = [DS * CI 4- (1. - DS) * ABL / (VBL * PB)] * 24450. / 202.51

       the parameters for the model are shown in Table 1.  Physiological parameters for mouse, rat,
and human were developed from literature data collected (Stan Lindstedt, Northern Arizona
University, personal communication) as part of an ongoing Physiological Parameters Work Group
effort sponsored by the ILSI Risk Science Institute .  Physiological parameters and partition
coefficients for the rhesus monkey were adapted from Crank and Vinegar (1992).  Partition
coefficients for the rat and mouse were taken from Jepson et al. (1985). Partition coefficients for
humans were assumed to be the same as for monkeys.  Metabolism was modeled as a first-order
process, scaled allometrically from the value determined in rats (Jepson et al., 1985). The model was
written in the Advanced Continuous Simulation Language (ACSL, Mitchell and Gauthier, Boston
MA) and was compared with experimental data using SimuSolv (Dow Chemical Co., Midland MI).
The Monte Carlo analysis was performed on the ACSL model with PBPK_SIM (K.S. Crump Div.,
ICF Int., Ruston LA).

-------
        Figure 2 shows the results of gas uptake analysis of CPFB in rats (Jepson et al.,  1985).  In
the gas uptake analysis, several animals are maintained in a closed chamber, and the air is
continuously recirculated. Oxygen is replenished and carbon dioxide is scrubbed as necessary to
maintain stasis.  A known, amount of a volatile chemical is  then added to the chamber, and the
concentration of the chemical in the chamber is monitored over time.  The rapid initial decline in the
chamber concentration of CPFB seen in Figure 2 is due to uptake by the animals' tissues and
demonstrates that CPFB is readily absorbed. Following the tissue uptake phase, any further decline
in chamber concentration would indicate loss of chemical due to metabolism.  The fact that the
concentration curve for CPFB almost levels out after the first few hours reflects the.fact that CPFB is
not extensively metabolized. By way of comparison, the chamber concentration of a more rapidly
metabolized chemical, bromopentafluorobenzene, decreased by more than 20% between hours 3 and 6
under the same conditions.  Using the PBPK model for CPFB, the closed chamber data was analyzed
to quantify the rate of metabolism. It was determined that metabolism was first-order, with a rate
constant of 2/hr (scaled to a 1 kg animal  by body weight to the -0.25 power).

Model Validation

        As  a  test of the model, a study was simulated in which rats were exposed 6  hours per day by
inhalation for 21 days to CPFB  at 30, 100, and 300 ppm (Kinkead et al., 1990a). Figure 3 shown the
measured and simulated venous  blood concentration of CPFB for the eleventh day of the exposure.
As a further evaluation of the model, inhalation exposures to CPFB on eight anesthetized rhesus
monkeys (Crank and Vinegar, 1992) were simulated.  In these experiments CPFB concentrations in
expired breath were measured during and after 15 minute exposures at 300 ppm.  The PBPK model
was evaluated in terms of its ability to relate exposure concentration and exhaled-air concentrations.
The results are shown in Figure 4.

                              TOXICOLOGICAL EVALUATION

       In order to assure that CPFB could safely be used as an intake simulant, a number of studies
were performed to evaluate its potential toxicity. These studies  were designed to elucidate any
short-term or long-term effects,  and to assess the likelihood that CPFB could be carcinogenic or
teratogenic.
Acute Toxicitv

-------
       Hie primary irritation hazard, sensitization potential, and acute inhalation toxicity of CPFB
were evaluated by Kinkead et al (1987).  CPFB demonstrated no potential for skin sensitization in
tests on guinea pigs, and was only a mild skin and eye irritant in rabbits.  Short-term exposure to
CPFB vapor poses no serious hazard by the inhalation route as all rats survived a 4-hour exposure to
an upper limit concentration of 4.84 mg/L (581 ppm), a concentration many orders of magnitude
higher than that which is likely to be encountered in the field.  Similarly, oral dosing indicates an
LD* of greater  than 5 g/kg,  which would classify CPFB as "practically non-toxic"  (Kinkead et al.,
 1990b).

 Mutagenicitv/Genotoxicitv

        CPFB was tested for potential genotoxic activity by three different laboratories CTu et al.,   .
 1986; Steele, 1987; Kutzman et al., 1990) using  a battery of in vitro assays (Table 2).  The first
 attempt to perform these assays CTu et al., 1986) was compromised by experimental difficulties
 associated with the tendency of CPFB to precipitate out of solution and to dissolve the dishes. In the
 second study (Steele,  1987), it was again noted that CPFB dissolved the standard plastic dishes, so the
 study was performed  in specially designed glass  dishes. A third study (Kutzman et al:, 1990) was
 performed by a reference laboratory since the results of the first two studies  seemed to be somewhat
 equivocal.

        .CPFB  does not appear to be mutagenic.   The Ames  Salmonella reverse mutation assay was
 uniformly negative in all studies, both with  and  without the  addition of a rat liver S9 metabolic
 activation system.  Similarly, all of the laboratories obtained negative results when CPFB was tested
  in mammalian cell culture for mutagenic activity at the HGPRT locus in Chinese hamster ovary cells.

         The results of tests for genotoxicity were less consistent. There was some evidence of
  CPFB-induced sister chromatid exchange and/or chromosomal aberration in the earlier studies,  but the
  final study detected no increases in chromosomal aberrations and only observed sister chromatid
  exchange with the addition of liver S9 metabolic activation (suggesting that  generation of significant
  levels of metabolite may be required to observe this effect). In the case of the assay for unscheduled
  DNA repair synthesis in primary rat hepatocytes, the first study suggested that CPFB produced
   increased repair of DNA damage; however, both the second and third studies failed to confirm this
   finding.  Cell transformation results were also variable, with only the second study showing any

-------
 indication of an ability of CPFB to induce morphological transformation in vitro in BALB/C-3T3
 cells.
        To resolve the question of whether CPFB could act as a genotoxic or cytotoxic agent under
 in vivo conditions, a 21-day exposure of mice to CPFB at 30, 100, and 300 ppm was performed
 (Kinkead et al., 1989).  Under these conditions CPFB did not induce an increase in sister chromatid
 exchange in the bone marrow of the exposed mice, and the rate of cellular proliferation in the bone
 marrow was not altered. Similarly, assessment of the micronucleated polychromatic and
 normochromatic erythrocyte populations during the exposures indicated a general absence of
 genotoxic activity.  A PBPK  model for CPFB was used to  assess the tissue exposure to CPFB during
 this study (Kinkead et al., 1990a). Based on the modeling, bone marrow tissue exposure to CPFB
 during the in vivo study was similar to or greater than the concentrations used in  the in vitro assays.

        The PBPK model described in this paper was used to reconfirm the results of this earlier
 analysis in the particular case of sister chromatid exchange.  A dose-related increase in sister
 chromatid exchange was observed -for 2-hr in vitro exposures to CPFB ranging from 100 to 250 mg/L
 (area under the curve ranging from 200  to 500 mg/L-hr) in the presence of metabolic activation.  For
 the in vivo study, bone marrow exposure to CPFB (as estimated by the model described in the
 Appendix) averaged 288 mg/L during the daily 6-hr inhalation exposures to 300 ppm CPFB, with a
 daily area under the curve in the marrow of 2023 mg/L-hrs. The lack of in vivo  response appears
 therefore to reflect differences between the  in vivo and in virro situation rather than failure to  achieve
 sufficient tissue exposure levels.   It is possible that the bone marrow does not possess sufficient
 metabolic activity, in comparison  with the in vitro situation, to generate the active chemical species.

        Full evaluation of the potential for CPFB to be carcinogenic would require a lifetime animal
bioassay. However, a reasonable assessment of the likelihood that CPFB could act as a carcinogen.
can be made on  the basis of the above results, taken together with the rather unremarkable results of
the subchronic exposures.  CPFB  does not appear to be mutagenic, either in the presence or absence
of metabolic activation,  and the questionable in vitro suggestions of genotoxicity were not born out by
the in vivo studies.  In addition, subchronic exposure (Kinkead et al., 1990b) did  not produce any of
the tissue changes, such as peroxisomal proliferation, which typically accompany  promotional
carcinogenesis in rodents.  Therefore, it is not likely that CPFB  would be carcinogenic, even under
the conditions of a lifetime bioassay.

-------
frihacute and Suhchronic Toxicitv

       Repeated exposure of rats to high concentrations of CPFB produced lethargy and
incoordination (1000 ppm, 6 hours/day, 4 days) or unresponsiveness (500 ppm, 6 hours/day,
15 days), but no tissue pathology (Gage, 1970).  No behavioral or historical effects were observed
for exposure to 250 ppm, 6 hours/day, for 15 days (Gage, 1970). [Note: Gage (1970) incorrectly
shows the concentration of the lowest exposure level as 50 ppm;  the original ICI report, TR/449,
records the concentration as 250 ppm - J.C. Gage, personal -communication.]

        In a more recent study (Kinkead et al., 1989), ten Fischer-344 rats and six B6C3F1 mice of
 each sex were exposed to 30, 100, and 300 ppm CPFB for 3 weeks (15 exposures).  Exposure to the
 highest concentration caused a reduction in the growth rate of rats, but did not affect the growth rate
 of mice. Both rats and mice showed a dose related increase in liver to body weight ratios.  Mice
 showed clear evidence of liver toxicity (hepatocytomegaly and hypertrophy) at the highest exposure
 concentration.  Another treatment-related change in the  livers of male and female mice and female
 rats was an increase in the incidence of single-cell necrosis in all CPFB-exposed groups.  The
 formation of hyaline droplets in the kidneys of male rats was also noted, but the severity of the lesion-
 was minimal, and no  other kidney effects were seen. Consistent with the earlier study, no behavioral
 effects were noted, even at the highest dose.

         In order to better evaluate the impact of prolonged or repeated exposure to  CPFB, as well  as
  to determine a no-observable-effect level,  a 13-week exposure of rats and mice was carried out at
  concentrations of 1.2, 6, and 30 ppm (Kinkead et al., 1990b; 1991).  No treatment-related effects
  were observed at any concentration in either  species.  In particular, the single cell necrosis seen in the
  3-week study at 30 ppm was not observed in the 13-week study at the same concentration. A review
  of the tissues from the earlier study confirmed the finding of an increase over control, but both the
  number and severity of the lesions were so slight that ifwas felt the finding was biologically
  unimportant.  Thus the only adverse effects seen were those noted for the 300 ppm exposure
  concentration in the  3-week study.  A concentration of 30 ppm was therefore recommended by the
  investigators as a no-effect level in humans to protect  individuals subjected to repeated inhalation of
  CPFB for extended periods.

   Reproductive Toxicitv

-------
       To evaluate the teratogenic potential of CPFB, time-mated Sprague Dawley rats were dosed
orally at 0.3, 1.05, and 3.0 g/kg/day on days 6 through 15 of pregnancy (Copper and Jarnot, 1992).
There was a significant reduction in maternal body weight and a significant increase in maternal liver
weight at the highest dose.  The percentage of post-implantation fetal loss was also greater only at the
highest dose.  Fetal weight and length differed significantly from the controls at both the high and
intermediate doses, indicating a slightly increased fetotoxicity compared to the dam.  The number of
malformations and variations observed at any of the doses did not differ from controls, suggesting
that CPFB is not teratogenic.

Metabolism

       Studies of the uptake of CPFB in a closed, recirculated chamber were consistent with a slow
rate of first order metabolism (Jepson et al., 1985).  In the same studies, the rate of metabolism of
the related compound, bromopentafluorobenzene, was unaffected by pretreatment with the potent
P450 inhibitor, pyrazole, suggesting that metabolism of these two compounds is not associated with
the mixed function oxidase system.  This finding contrasts with the metabolism of the related
compound, hexachlorobenzene (HCB), which is  characterized by both an oxidative (P450) pathway
and a glutathione conjugation (GST) pathway (Renner, 1988).  This apparent difference between
CPFB and HCB is consistent with the results of a comparative study of a series of dihalomethanes
(Gargas et al., 1986), which also feature competitive P450 and GST metabolism.  This study
demonstrated that the fluorine-substituted congeners, CH2F2 and  CH2FC1, showed little evidence of
P450 activity, whereas compounds containing chlorine and/or bromine, but not fluorine, were readily
metabolized by both pathways.  Of course, these results were observed in rodents,  and the possibility
of species differences in the metabolism of CPFB cannot be ruled out.  Evaluation of CPFB
metabolism in human tissues would be necessary to confirm the assumption of equivalent metabolism
across species.
       In the case of HCB, the GST pathway initially produces N-acetyl cysteine conjugates which
cleave to form chlorothiophenols, which are in turn subject to further metabolism (Renner, 1988). It
can therefore be hypothesized that the liver toxicity associated with repeated exposure to CPFB may
result from the generation of the analogous metabolite, pentafluorothiophenol (PFTP), a toxic
compound with an LD^, of 56 mg/kg (NIOSH, 1992).

-------
                        EXPOSURE GUIDELINE DETERMINATION

       The critical effect for evaluation of safe exposure to CPFB is the liver toxicity associated with
repeated exposure (Kinkead et al., 1990a). Specifically, hepatocytomegaly and hypertrophy were
observed in mice following exposure to 300 ppm CPFB, 6 hours/day, for 3 weeks, and the liver to
body weight ratio in rats and female mice were increased in a dose related fashion.  Increased single
cell necrosis was also observed at 30 ppm and 100 ppm in the same study, but this effect  was not
considered theologically significant, and neither the necrosis nor the increased liver to body weight
ratio were reproduced in a subsequent study at 30 ppm for 13 weeks (Kinkead et al, 1991).  In the
traditional approach, taking  30 ppm as a No Observed Adverse Effect Level (NOAEL),- adjusting for
the difference in daily exposure duration  (6 hrs for animal studies, 8 hrs for humans), and dividing by
a factor of 33 to provide a margin of safety, yields a recommended exposure guideline of 0.7 ppm for
a daily (8-hour) time-weighted average.

        The rationale for the factor of 33 used in the traditional guideline calculation is as follows.
 First, the animal NOAEL must be adjusted for the relationship between the duration of exposure in
 the animal study and the anticipated duration of  exposure  in the human scenario. One aspect of this
 adjustment is described above: adjusting for the  difference in daily exposure duration. Assuming a
 maximum daily exposure duration of 8 hrs when simulant training is performed, the adjusted NOAEL
 is 30 * 6/8 = 22.5 ppm. However, the actual anticipated human exposures are brief and infrequent,
 associated with special training exercises which  are not expected to be a common occurrence.
 Therefore the 3-week and 13-week rodent studies represent much more prolonged exposures than the
 human exposure scenario, with less opportunity for recovery between exposures. It is common
 practice to apply factors of up to 10 to extrapolate from short-term to longer-term toxicity (Dourson
 and Stara, 1983). 'in this case, the extrapolation is in the other direction, from relatively long-term to
 shorter-term, so an  inverse factor is justified. To be conservative, a factor of 1/3 was selected,
  yielding an adjusted NOAEL of 22.5 / 0.33  =  67.5.  The traditional guideline then applies a safety,
  or uncertainty, factor  of 100, with one factor of 10 to account for uncertainty in die extrapolation
  from animal to man and a second factor of  10 to account for human variability, resulting in the
  guideline of 0.7 ppm.

  The selection of 100 as the safety factor to  be  applied in this case follows a convention which,
   although basically empirical, can be at  least partially justified  on the basis of quantitative
   pharmacokinetic principles (Dourson and Stara, 1983).   For example, the factor often usually applied

-------
 for extrapolation from animals to man reflects a conventional wisdom based primarily on experience
 with the more common exposure routes, oral and intravenous, for chemicals which are themselves
 toxic and are cleared or detoxified by processes  which scale roughly with surface area.rather than
 body weight. Under these conditions, pharmacokinetic and empirical allometric considerations justify
 such a factor for rodent-to-human extrapolation based on the relationship between applied dose and
 tissue exposure  (area under the concentration-time curve) as a function of body weight (NRC, 1986).
 However, for inhalation exposure to a volatile, poorly soluble chemical such as CPFB, these same
 principles lead to an expectation of similar area under the curve (AUC) of both parent and metabolites
 for equivalent external concentration and exposure duration (that is, for equal time-weighted average
 concentrations).  In order to make quantitative use of these pharmacokinetic principles, the model
 described in this paper was  used to calculate the daily AUC in the liver for exposure of rats and
 humans to 30 ppm CPFB  for 6 hrs.  The AUC in the liver predicted for rats was 46 mg/L-hrs,
 whereas for humans,  even under conditions of moderate exercise, it was 72 mg/L-hrs, a difference of
 less  than a factor of two.  Thus the usual animal to human extrapolation factor of 10 is not justified  in
 this  instance.

        The second factor of 10, which accounts for  human heterogeneity, would similarly be
 susceptible to quantitative evaluation if the distribution of human susceptibilities could be estimated.
 For  a specific chemical  toxicity, the variance of the response distribution in the human population  will
 depend on the steepness of the chemical dose/response curve, which can be determined from animal
 studies, and on the extent  of variability in the human population of the pharmacokinetic and
 pharmacodynamic parameters mediating the response (Hattis, et al., 1987).  Coupling of Monte Carlo
 analysis and PBPK modeling provides a method for directly estimating the impact of parameter
 variation on risk (Clewell, 1993).
        A pharmacokinetically-driven guideline calculation is based on calculation of equivalent
effective tissue doses for the animal and human scenarios.  In the case of the liver toxicity associated
with prolonged exposure, the AUC for CPFB in the liver was selected as the appropriate tissue dose.
The AUC is generally regarded as an appropriate dose surrogate for cumulative, reversible toxicity,
such as that seen with CPFB.  The daily AUC in the liver calculated by the model at the rodent
subchronic NOAEL of 30 ppm was 46 mg/L-hrs.  The selection of an appropriate uncertainty/safety
factor for the pharmacokinetic approach was based on three considerations.  First, it was determined
that there was not yet sufficient information on the variability of human susceptibility to liver toxicity
to permit calculation of a more accurate substitute for the default uncertainty factor of 10, so the

                                               10

-------
default value was used.  Second, based on Monte Carlo analysis of the impact of parameter
uncertainty on the dose surrogate predictions of the CPFB model, an uncertainty factor of 2.3 was
included for animal-to-human extrapolation.  This factor represents uncertainty in the accuracy of the
PBPK model predictions, as distinguished from the default factor of 10, which represents total
uncertainty in  the animal to human extrapolation when pharmacokinetics is not considered.  The
major contributor to the PBPK uncertainty factor is the lack of data on the human metabolic capability
for CPFB.  Additional data, e.g. from in vitro metabolism studies on human liver tissue, would
greatly reduce this uncertainty factor.

        The final consideration for the overall uncertainty factor was  the infrequent nature of the
 anticipated human exposure, as discussed for the traditional guideline.  A factor of 1/3 was again
 used. Thus the overall safety factor for the pharmacokinetic guideline is 2.3  * 10 * 1/3 - 7.7. The
 model  was  therefore exercised to predict the exposure concentration at  which the AUC in the liver for
 a human would be one-eighth of the value in the animal at the NOAEL.  For an 8-hr time-weighted
 average exposure at 1.8 ppm, the calculated AUC in the liver was 5.6  mg/L-hrs, a factor of 8 below
 that at the animal subchronic NOAEL. The pharmacokinetically based guideline of 1.8 ppm is  more
 than a factor  of two higher than the traditionally derived guideline.

         In addition to liver toxicity, there is limited evidence (Gage, 1970) of behavioral effects at
 higher CPFB concentrations (500 to 1000 ppm).  Any behavioral deficit induced by CPFB could not
 only degrade performance during a training exercise, but could also  increase the likelihood of
 subsequent exposure through improper use of protective gear.  To avoid any behavioral effects
 potentially associated with brief exposure to higher concentrations, a short-term guideline was also
  developed. The traditional guideline is a 3 ppm ceiling limit based on the 300 ppm NOAEL for
  behavioral effects (Kinkead et al., 1991),  with a safety factor of 100.  The rationale for this guideline
  is the same as for the liver toxicity except that no exposure duration adjustment is required since this
  is a ceiling limit.  In the case of acute behavioral effects, toxicity generally appears to be correlated
  with peak concentration rather than area under the curve. The model predicts peak blood
  concentrations of 24.6 mg/L and 26.6 mg/L in rats and mice, respectively, at the acute NOAEL of
  300 ppm. Providing a margin of safety of 1.8 for PBPK uncertainty and 10 for interindividual
   variability, the model determined that an exposure of 31 ppm CPFB would produce a peak blood
   level of 1.4 mg/L in humans. This pharmacokinetically derived ceiling is roughly a factor of ten
   higher than the traditionally derived value.
                                                  11

-------
        Finally, fetotoxic effects were observed in rats dosed orally, with a NOAEL of 300
mg/kg/day (Cooper and Jarnot, 1992). The traditional guideline calculation requires a dose-route
adjustment from the oral route used in the animal study to the inhalation route of concern for human
exposure.  The default calculation equates routes on a mg/kg basis, assuming an inhalation rate of 10
cu.m per 8 hrs:
                              300 mg/kg * 70 kg /  10 cu.m.  = 2100 mg/cu.m
                              2100 mg/cu.m. * 24.45 cu.m./mole  / 202.5 g/mole = 254 ppm

Using safety factors of 10 for animal-to-human extrapolation uncertainty,  10 for oral to inhalation
extrapolation uncertainty, and 10  for human variability results in a guideline of 0.25 ppm.

        In the pharmacokinetic approach, the oral exposure can be used to develop an inhalation
guideline by using the PBPK model to estimate the peak blood concentrations and AUC for CPFB  in
the oral rodent study and comparing them with those  achieved during human inhalation exposures.
Both peak concentration and AUC are evaluated as dose surrogates, since the mechanism of
fetotoxicity in this case is not established. For an oral dose of 300 mg/kg in the rat, the model
estimates a peak blood level of  123  mg/L and an area under  the blood curve of 109.2 mg/L-hrs. In
this case, PBPK uncertainty factors  of 4.8 and 5.1, respectively,  are required, due to additional
uncertainty from the oral uptake parameters.  Taken together with a factor of 10 for human
variability, the target dose surrogates in the human are a peak blood level of 2.56 mg/L and an AUC
of 2.14 mg/L-hrs. For human1 inhalation, the peak blood  level  is predicted to be 2.5 mg/L at 56
                                                             6
ppm, , while the area under the blood curve for an 8  hr exposure is 2.04 mg/L-hrs at 5.2 ppm. Thus
the pharmacokinetically-derived inhalation guideline for the prevention of fetotoxic effects is 5 ppm,
based on AUC in the blood. This guideline is a factor of 20 above the traditionally-derived guideline.

                                        CONCLUSION
       CPFB possesses a remarkable combination of properties, making it an attractive candidate for
use as an intake simulant in chemical defense field training exercises. It is volatile and unreactive,
simplifying dissemination, and mimics the performance of typical vapor threats in terms of persistence
and canister penetration.  It does not appear that CPFB would present any significant health hazards
to personnel under the envisioned use.  A thorough lexicological evaluation indicates that CPFB is not
acutely toxic or teratogenic and is not likely to be carcinogenic.  Chronic liver toxicity was observed
only after prolonged exposure to high concentrations.  Based on a pharmacokinetic analysis,  it is

                                               12

-------
recommended that field exercises be designed to avoid short-term exposures to concentrations greater
than 30 ppm, with the daily (8-hour) time-weighted average not to exceed 2 ppm. Since field
analytical methods can measure CPFB at part per billion levels, this should not be an impediment to
its use in training exercises.  By comparison, a traditional approach to guideline generation would
result in a short-term limit of 3 ppm with an 8-hr time-weighted average of 0.25  ppm.

                                        REFERENCES

Barnes, D.G., and M.L. Dourson (1988) "Reference dose *icnl.  Pharmacol.. 3, p. 234-228.

  Gage, J.C. (1970)  "The subacute inhalation toxicity of 109 industrial chemicals,"  Brit. J. Induste,  .
  Med.. 27, p. 1-18.

  Gargas, M.L., Clewell, HJ., and M.E.  Andersen (1986)  "Metabolism of dihalomethanes in vivo:
  Differentiation of kinetic constants for two independent pathways," lojdc^ApjzLJ^rma^ 82, p.
  211-223.
                                                 13

-------
 Hattis, D., Erdreich, L., and M. Ballew (1987) "Human variability in susceptibility to toxic
 chemicals - a preliminary analysis of pharmacokinetic data from normal volunteers." Risk Analysis.
 7:4, p. 415-426.

 Jepson, G.W., H.J. Clewell, III, and ME. Andersen (1985) "A rapid, physiologically based method
 for evaluating candidate chemical warfare agent uptake simulants," AAMRL-TR-85-045, Armstrong
 Aerospace Medical Research Laboratory, Wright-Patterson Air Force Base, Ohio.

 Kinkead, E.R., W.J. Bashe, D.M. Brown, and S.S. Henry (1987)  "Evaluation of the inhalation
 toxicity and sensitization potential of chloropentafluorobenzene," in 1986 Toxic Hazards Research
 Unit Annual Report, AAMRL-TR-87-020, NMRI-87-2, Armstrong Aerospace Medical Research
 Laboratory, Wright-Patterson Air Force Base, Ohio, p.  131-135.

 Kinkead, E.R., B.T. Culpepper, H.G. Wall, R.S. Kutzman, C.D. Flemming, C.J. Hixon, and R.R.
 Tice (1989) "Evaluation of the potential of inhaled chloropentafluorobenzene to induce toxicity in
 F-344 rats and B6C3F1 mice and sister chromatid exchanges and micronuclei formation in B6C3F1
 mice," AAMRL- TR-89-037, Armstrong Aerospace Medical Research Laboratory, Wright-Patterson
 Air Force Base, Ohio.

 Kinkead, E.R., H.G. Wall, C.J. Hixson, R.R. Tice, R.S. Kutzman, and A. Vinegar (1990a)
 "Chloropentafluorobenzene: short-term inhalation toxicity, genotoxicity and physiologically-based
pharmacokinetic model development," Toxicol. Indust. Health.  6, 6, p. 533-550.

Kinkead, E.R., S.K. Bunger, E.G. Kimmel, C.D. Flemming, H.G. Wall, and J.H. Grabau (1990b)
 "Effects of a 13-week chloropentafluorobenzene inhalation exposure of Fischer 344 rats and B6C3F1
mice,"  AAMRL-TR-90-064, Armstrong Aerospace Medical Research Laboratory, Wright-Patterson.
Air Force Base, Ohio.

Kinkead, E.R., S.K. Bunger, E.G. Kimmel, C.D. Flemming, H.G. Wall, and J.H. Grabau (1991)
"Effects of a 13-week chloropentafluorobenzene inhalation exposure of Fischer  344 rats and B6C3F1
mice," Toxicol. Indust. Health. 7, 4, p.309-318.
                                             14

-------
Kutzman, R.S., B.C. Myhr, T.E. Lawlor, D.C. Valentine, R.R. Young, H. Murli, M.A. Cifone, and
B M Jarnot (1990) "Genetic toxicity assessment of chloropentafluorobenzene," AAMRL-TR-90-048,
Armstrong Aerospace Medical Research Laboratory, Wright-Patterson Air Force Base, Ohio.

NIOSH (1992)  *»pton, nf Toxic Eff^ nf rhamlcal Substances, U.S. Department of Health and
Human Services,  Washington D.C.

National Research Counci, (1986) MpUn, W^ «* H;alft. Vo,. 6, d, 6, p. 193-200, NaUona,
Academy Press, Washington D.C.

 Ramsey J.C. and M.E. Andersen (1984)  "A physiologically based description of the inhalation
 pharmacokinetics of styrene in humans and rats," Tojdc^^El^harmacoL, 73, p. 159-175.   .

 Renner, G. (1988)  "Hexachlorobenzene and its metabolism," Tovirol  Fnviron. Chem., 18, 1,
 p.51-78.

 Steele, V. (1987) "Biological activity of chloropentafluorobenzene," AAMRL-TR-87-039, Armstrong
 Aerospace Medical Research Laboratory, Wright-Patterson Air Force Base, Ohio.

 Tu  A M G Broome and A. Sivak (1986) "Evaluation of chloropentafluorobenzene in a battery of
  in litre short term assays," AAMRL-TR-86-003, Armstrong Aerospace Medical Research Laboratory,
  Wright-Patterson Air Force Base, Ohio.

  Vinegar  A   D.W. Winsett, M.E. Andersen, and R.B. Conolly (1990) "Use of a physiologically
  based pharmacokinetic model and computer simulation for retrospective assessment of exposure to
  volatile toxicants," Tnhal. ToxicoU. 2, p.  119-128.
                                                15

-------
Table 1: PBPK Model Parameters




  UNSCALED PARAMETERS

BW
KA
ALVS
DS
QCC
QPC

Body Weight (kg)
Oral Uptake Rate (/hr)
Alveolar Dead Space (Fraction)
Bronchiolar Dead Space (Fraction)
Cardiac Output (L/hr, 1 kg animal)
Alveolar Ventilation (L/hr, 1 kg
Mouse
0.023
5.0
0.0
0.3
16.5
29.0
Rat
0.22
5.0
0.0
0.3
11.6
21.2
Monkey
8.7
5.0
0.4
0.45
12.0
17.0
Human
70.0
5.0
0.0
0.3
18.0
35.0
Tissue Blood Flows (Fraction of Cardiac Output): ,
QFC
QGC
QLC
QMC
QRC
QSC
Flow to Fat
Flow to GI Tract
Flow to Liver
Flow to Bone Marrow
Flow to Rapidly Perfused Tissues
Flow to Slowly Perfused Tissues
0.030
0.166
0.036
0.110
0.409
0.249
0.058
0.183
0.032
0.110
0.362
0.255
0.052
0.185
0.065
0.110
0.348
0.240
0.052
0.185
0.065
0.110
0.348
0.240
Tissue Volumes (Fraction of Body Weight):
VBL
FVC
VGC
VLC
VMC
VRC
VSC
Volume of Blood
Volume of .Fat
Volume of GI Tract
Volume of Liver
Volume of Bone Marrow
Volume of Rapidly Perfused Tissues
Volume of Slowly Perfused Tissues
0.070
0.100
0.033
0.050
0.030
0.041
0.550
0.070
0.070
0.033
0.040
0.030
0.020
0.600
0.070
0.190
0.045
0.027
0.020
0.026
0.709
0.070
0.050
0.045
0.027
0.020
0.026
0.569
            16

-------
                      Table 1 (continued):  PBPK Model Parameters
Partition Coefficients:
PB
PF
PG
GI Tract/Blood
            Bone Marrow/Blood
            Richly Perfused Tissue/Blood
            Slowly Perfused Tissue/Blood
            =3===
 Metabolic Parameters:
 KFC
 Rate Constant for 1st Order Pathway
 (/hr -  1 kg animal)
             Affinity of Saturable Pathway (mg/L)
             Maximum Velocity of Saturable
             Pathway (mg/hr, 1 kg animal)
                                             17

-------
                     Table 1 (continued): PBPK Model Parameters
                              SCALED PARAMETERS
QC = QCC*BW**0.75
QP = QPC*BW**0.75

QF = QFC*QC
QG = QGC*QC
QL = QLC*QC
QM = QMC*QC
QR = QRC*QC
QS = QSC*QC

VBL = VBLC * BW
VF = VFC*BW
VG = VGC*BW
VL = VLC*BW
VM = VMC*BW
VR = VRC*BW
VS = VSC*BW

KF = KFC/BW**.25
VMAX = VMAXC*BW**0.75
                               DOSE SURROGATES

Amet        Total amount metabolized per unit body weight (mg/kg)
AUCB       Area under the curve of arterial blood concentration of CPFB (mg/L-hrs)
AUCL       Area under the curve of liver concentration of CPFB (mg/L-hrs)
AUCM      Area under the curve of CPFB in the bone marrow (mg/L hrs)
CA         Concentration of CPFB in the arterial blood (mg/L)
CL         Concentration of CPFB in the liver (mg./L)
CM   ,      Concentration of CPFB in the bone marrow  (mg/L)
CV         Mixed venous blood concentration of CPFB  (mg/L)
Dose        Total amount inhaled during exposure (mg/kg)
                   = integral of QP *  (CALV - CX) / BW
                                        18

-------
                  Table 2: Summary of In Vitro Results for CPFB
IN VITRO ASSAY
Ames Salmonella mutagenicity:
- S9 activation
1 + S9 activation
1 CHO/HGPRT gene mutation:
- S9 activation
1 + S9 activation
1 	 	
1 CHO sister chromatid exchange:
	 — 	 	 	
- S9 activation
| + S9 activation
1 CHO chromosome aberration:
	 	 	 	 	
1 - S9 activation
1 + S9 activation
L 	 — — 	 	 	
1 Primarv rat hepatocyte DNA repair
|j BALB/C-3T3 cell transformation:
|| - S9 activation
II + S9 activation
7
Tu et al.

-


-
-

-
-

+/-
+ /-
+ /-

-
1 a
J 	 _ 	
1"
Steele

-
	 -

-
-

+/-
+ /-
—

+ /-
'"—
+ /-
-

1 + /-
1 +
Kutzman et al.

-
-

-
-

-
+

-
-
-


	 ~
Not reported.
                                         19

-------
Figure 1

Figure 2



Figure 3



Figure 4
                      FIGURE CAPTIONS

Diagram of the physiologically based pharmacokinetic model of CPFB.

Computer simulation (solid line) vs. observed (x) chamber concentrations (ppm) over
time (hours) for  rats exposed to initial concentrations of 10, 250, 1000,  and 1800
ppm CPFB in a closed, recirculating chamber system. (Reproduced from reference 4)

Model-predicted (lines) and measured  (boxes) venous blood concentrations on 11th
day of exposure of rats to CPFB for 6 hours per day at 30 ppm (a), 100 ppm (b), and
300 ppm (c).

Model predicted (lines) and measured  (points) exhaled air concentrations  for rhesus
monkeys exposed to 300 ppm CPFB for  17 - 20 minutes.
                                            20

-------
           QP'CI
QOCV
  LUNG
   GAS
EXCHANGE
     QR-CVR
     QM*CVM
 RAPIDLY
PERFUSED
 TISSUES
                 MARROW
      QS*CVS
    I  QF'CVF
  SLOWLY
PERFUSED
  TISSUES
                   FAT
                 Gl TRACT
                                  QC'CA
                             QR*CA
                             QM-CA
                             QS*CA
                             QF*CA
                             !QG*CA
              QG'CVG
  (QL+QG)«CVL
                   LIVER
            _|O=OR!NK


            1  QL-CA
               MFO ty ty GST
                   Figure 1

-------
                      Wdd U3SWVHO 038010 Nl ONOO) - dO
                                   Figure 2
_

-------
  4-
 .3-
  °0       10      20-30      40.     50
  10
3
i
3
S
 i
a
            10      20   T   30       40      SO
10-
8-
s"
4-
2-

c

a ' : :
X9 ' ! i !
7 	 ; '
m-
~^^" i
10 20 T 30 40 SC
               Figure  3

-------
      1
     g
                                      rs

                        S   8    g
                    IW«O«o *~
                      S   8
                    HMAO-o "
                                      §
Figure k

-------
          BIOGRAPHICAL SKETCH:  DR. MICHAEL L. DOURSON




located in Cincinnati, Ohio.  ECAO is a riSK-°~'°   ntfict> of Health
the D.S. Environmental Protection agency's (EPA-s) _office "^J^
and  Environmental  assessment.   Dr.  to^s°* >£* J^1* ^^ °f

i!vrdfffPerSrorXEs?'-°ohatrli%^rofSl^^^^

 1                           9Sriea^ip                ?

                                tjr-s^^SiS ^ss
?r a pSst  pSSident  of the Society  of  Toxicology's Specialty
Sl^ioS  £1 ^k  Assessment.    Dr   Dourson ff.^^^^3
publications and presentations in the  area  of risk  assessment,
primarily for the assessment of noncancer health effects.

-------
                    RISK ABOVE THE REFERENCE DOSE (RfDVBENCHMARK DOSE (BMD)
                                   Michael Dourson and Richard Hertzberg

               Abstract
                    Current methods to estimate noncancer health risk are  limited.   Situations exist
               where subthreshold doses such as Reference Doses (RfDs) or Reference Concentrations
               (RfCs) are exceeded and little is known about the possible health risk. A recent model
               indicates that toxicity data viewed  as categories of pathology has potential for exploring
               such risk.  What appear to be reasonable estimates of rjsk above the RfC are found with
               toxicity data for manganese.
                   This model is also compared to another new approach  - the benchmark dose for
               several chemicals.  Differences between these two approaches are briefly discussed.

               Introduction
                   The RfD has been the  mainstay of noncancer risk assessment in the U.S. EPA for
               several years. The RfD is defined as: an estimate (with uncertainty spanning perhaps an
               order of  magnitude) of a daily exposure to the human population  (including sensitive
               subgroups) that is likely to be without an appreciable risk of deleterious effects during a
               lifetime.
                   As displayed in Figure 1, much interest exists in the estimation of health risk above
              some level, such as an RfD  or RfC. Little progress has been made, however, primarily
              due to the multiplicity of effects, the changing severity and intensity of individual effects
              as dose increases, and the lack of mathematical tools.
                                                   -1-
_

-------
     The purpose of this manuscript is to present and analyze toxicity data for several
chemicals with a method of Hertzberg (1989) that addressees some of these issues.
Risks above the RfD or RfC are estimated with this model.
     In addition, recent interest has been expressed in replacing the NOAEL-based RfD
with a benchmark dose (BMD): a statistically-derived lower confidence limit on a dose
associated with a specified level of excess risk such as 1, 5 or 10%.  The arguments
favoring  the  BMD  are that  statistical  models can  be used,  and  that a consistent
interpretation can be made of the BMD across studies and across chemicals. We show
briefly in this paper (and more extensively elsewhere: Hertzberg and Dourson, 1993) that
the BMD is not consistent with the philosophy of the RfD, and that it leaves several issues
unresolved.

Methods
     We reviewed inhalation toxicity data for manganese and judged exposure or dose
groups as one of four very broad categories of toxicity; either no-observed effect, no-
observed-adverse effect, adverse effect, or frank effect.  We regressed these ordered
categories against both concentration (or dose) and exposure duration  using a logit
 model.  Ordered regression obviates pathologic "distances" among categories.
      Based on an analysis of all data as shown schematically in Figure 2 for one study,
 it is possible to determine the probability of NOEL, NOAEL, LOAEL and PEL for a given
 chemical. This is shown hypothetical^ in Figure 3.  In mathematical terms, categorical
 regression can be seen as follows:
                                       -2-

-------
               RfD Definition
          "is likely to be"
          "without appreciable risk"
          "deleterious effect"
      Regression Model
P(*)>0.95
r<10'2
toxicity category = moderate or
lethal adverse effect
     This  leads  to a new  RfD definition:   P(r<10'2|dose 0.95, where r = P
 (severity>1).  We selected 10'2 risk and the 95%^confidence level only to illustrate the
 method. Standard values for these decision criteria have not been adopted by EPA. The
 value  10~2 is a more realistic goal than the  10'6 risk often used for carcinogenic risk
 because  most of the noncancer effects we consider are sublethal,  and many are
 reversible.

 Results
     Toxicity data for manganese were excerpted from available literature. These were
 available  incidence  data from human studies as  shown in Table 1.   The  resulting
 probability statements for manganese are interpretable as human incidence for either an
 adverse effect (e.g.,  finger tremor) or frank  effect  (e.g., disturbed gait) as shown  in
 Figure 4.  For most published toxicity studies, effects are noted only for the dose group,
 so the probability statements are likewise interpretable only at the dose group level (i.e.,
the probability that a dose group will have the effect).
Discussion
     The categorical regression model described by Hertzberg and Miller (1985) and later
papers (Hertzberg, 1988, 1992; Hertzberg and Wymer, 1991; Guth et al., 1991; Farland
                                       -3-

-------
and Dourson, 1991) is one approach that incorporates judgments of toxicity along with
response  rate  into a  statistical  characterization of the overall  exposure-response
relationship.  The model can be used  to estimate a BMD, or can  be used to estimate
toxicity risk at any exposure  level.  The "risk" is no longer a single number, but a vector
of numbers, one for each category.
     The categorical regression approach has two distinct advantages over the NOAEL-
RfD and BMD-RfD procedures described above.  First, the approach is easily adapted
to most types of toxicity data, from  judgments of overall severity of toxic effect for the
dose group to measured responses on each individual.  Second, all relevant toxicity data
are included in the regression. Third, a consequence of the second advantage, this goal
of this approach is highly consistent with that of the NOAEL-RfD method: to  produce
 regulatory information that incorporates all toxic effects. In particular, an exposure level
 can be estimated by regression that is quite similar in interpretation to the NOAEL-based
 RfD.  Fourth,  the judgmental step involves evaluation  of overall toxic impact on the
 exposed individual, allowing comparison across target  organs, and across chemicals
 when several organs are affected.
      However, the probabilities generated by  categorical regression are usually limited
 to whether or not  a dose or exposure  group, and not an individual, is at risk. When
 incidence data are used in the analysis (such as for manganese shown here), actual
 population risk estimates are possible.
      Perhaps the greatest advantage of categorical regression is that this method can
  compare the likely health risk above the RfD  or RfC for several chemicals.  In risk
  management decisions, such comparison is often necessary. Figure 5 demonstrates this
  concept hypothetical^.
                                         -4-

-------
 Disclaimer
      Although the research (or other work) described in this article has been funded
 wholly or in part by the United States Environmental Protection Agency, it has not been
 subjected to the Agency's required peer and administrative review and, therefore, does
 not necessarily reflect the view of the Agency.  No official endorsement should be
 inferred.

 References
 Farland, W.  and  M.L  Dourson.   1992.   Noncancer health endpoints: Approaches to
 quantitative risk assessment. (In press)

 Guth,  D.J., A.M.  Jarabek, L Wymer  and R.C.  Hetzberg.  1991.   Evaluation of risk
 assessment methods for short-term inhalation exposure. Presentation at the 84th Annual
 Meeting  of the Air & Waste Management Association, June 16-21, 1991, Vancouver,
 British Columbia.

 Hertzberg, R.C. 1988.  Studies on toxicity applicable to risk assessment (STARA) user's
 guide.    Quantitative  toxicity  data  and  graphics  on   environmental  chemicals.
 Environmental Criteria and Assessment Office, U.S. EPA, Cincinnati, OH.  February.

Hertzberg, R.C. 1989.  Fitting a model to categorical response data with application to
special extrapolation to toxicity. Health Phys. 57(Suppl. 1): 404-409.
                                      -5-

-------
Hertzberg, R.C.  1992.  Studies on toxicity of mixtures and interacting chemicals user's



guide.  (MIXTOX) U.S. EPA, Environmental Criteria and Assessment Office, Cincinnati,



OH 45268. Available on diskette.







Hertzberg,  R.C.  and M.L. Dourson.  1993.  Using categorical regression instead of a



NOAEL  to  characterize a toxicologist's judgment  in  noncancer risk assessment.




Submitted.







Hertzberg, R.C. and M. Miller. 1985.  A statistical model for species extrapolation using



categorical response data.  Toxicol. Ind. Health.  1: 43-57.







Hertzberg, R.C.  and  L. Wymer.  1991.  Modeling the severity of toxic effects.   In:



Proceedings papers from the 84th Annual Meeting and Exhibition of the Air & Waste



 Management Association, June 16-21,  1991, Vancouver, British Columbia.
                                        -6-

-------
                                                FOG OF
                                                UNCERTAINTY
                                                            o

                                                            
-------
                                SLIGHT BODY WEIGHT DECREASE
RfD

 I -
    NOEL  NOAEL   LOAEL
UFxMF
                                              FAT IN LIVER CELLS
                                              (CRITICAL EFFECT)
                                              CONVULSIONS
                                                ENZYME CHANGE
                                           PEL
                       FIGURE 2

 Typical judgments on doses used to determine RfDs in on© study.
                           -8-

-------
           For any chemical it is possible to....
                            NOEL
                         DOSE
     NOAEL
    DOSE
                      AEL
                          DOSE
FEL
                                                        DOSE
                                 FIGURE 3

     Probabilities of various effect or no effect levels, with dose based on a review of
all data.
                                    -9-

-------
CL
                                                        LLJ
                                                        a:
                                                             o

                                                             1
                                                             0)
                                                             o
                                                             o
                                                             o
                                                             O)
                                                             -a
                                                             1
                                                             3
                                                             IS
                                                             TJ

                                                             en
                                                             3-
                                                             t3
                                                             0)

                                                              o
                                                              E
                                                              Q)
                                                              0}
                                                              CD
                                                              O)
                                                              CD
                                                              8
                                                              1
                                                              JK
                                                              o
                                                              1
O CD
P
li
a. &
   O
   o
                          -10-

-------
                             CHEMICALS
                         ABC
                           RfDs
DOSE
                           1-FOLD         1OO-FOLD

                          MULTIPLE OF THE RFD
                                  FIGURE 5

     For multiple chemicals it is thus possible to compare the risk of unacceptable effect
at existing exposures above the RfD or RfC.
                                     -11-

-------
                       TABLE 1



Human Studies on Manganese Exposure and Resulting Toxicity
S3====— — — r-
Study
.
Bradawy and
Shakour, 1985
Chandra et al.,
1981



Cook et al.,
1974
Davies, 1946
Emara et al.,
1971
Flinn et al.,
1941
Lauwery et al.,
1985
Nogawa et al.,
1973
Rodier, 1955
Roels et al.,
1987
Saric et al.,
1977

•
Number
=====
30
35
30
20
20

20

6

40-124
36

34

85

1222

-4000
141
204
190
66
268
17
•18
1
Exposure (HEC)
(mg/m3) 	 I
	 I
0.36
1.1
2.5
0.11
0.20

0.63

—
0.89-4.0

0.072-4.8
2.4-15

11-32

0.34

0.004

91-160
0.35
0.00005-0.00007
0.0007-0.01 1
0.17-0.38
0.11-1.8
3.4-4.0
5.9-7.3
1.8-5.8
Duration
(days)
	 =L.
?
?
?
7370
7670

5150

365-6205

2920
365-5840

365-1095

2920

730

30-3650
2592
9
?
-4015
-4015
-4015
-4015
>7300
Effect
=—4
CMS
CMS
CMS
CMS and
pulmonary
CMS and
pulmonary
CNS and
pulmonary
CNS

pulmonary
CNS

CNS

reproductive

pulmonary

CNS
CNS and
pulmonary
none
CNS
none
CNS
CNS
CNS
CNS
Severity and
- Percentage
AEL, ?
AEL, ?
AEL, ?
AEL, 25 I
AEL, 50

AEL, 40

FEL, ?

PEL, 3 1
FEL, 17
	
FEL, 24
	 J

AEL, ?

AEL, ? !
	
FEL, 4
AEL, 15
NOEL, 100
AEL, 5.8
NOEL, 100
AEL, 20
AEL, 18
AEL, 28
FEL, 100
                       -12-

-------

Study
Schuler
et al., 1957
Smyth et al.,
1973
Tanaka and
Lieben, 1969
=====
Number
83
71
38
117
TABLE 1 (cont.)
Exposure (NEC)
(mg/m3)
1.8
4.8
0.00004ato<1.8
>1.8
Duration
(days)
2980
2920-9490
?
7
Effect
CNS
CNS
none .
CNS
•
.
Severity and
- Percentage
PEL, 1 1
PEL, 7
NOEL, 100
PEL, 6
aAmbient background
                                 -13-

-------
              BIOGRAPHICAL SKETCH: DALLAS M. HYDE, Ph.D.


              Dr. Hyde received his undergraduate degree from the University of California Irvine
              in 1 967; his Master's from Whittier College in 1 972; and his Ph.D. from the University
              of California, Davis in 1976.
              to the University of California, Davis in 1979.

              In 1985-1986, Dr. Hyde was the recipient of the National Research Service Award
              Senior Fellowship. NIH, in the laboratory of Dr. Peter Hensen, at the National Jewish
              Hospital in Denver, CO.

              Dr  Hyde's research  focus  has  included the role of  neutrophils in oxidant-induced
              e^thelia? Injury  and  repair, and cellular  and molecular mechamsms  of f.broblast
              proliferation and collagen synthesis in experimental pulmonary fibrosis.
I/-* •'

-------
                      Morphometry of Pulmonary Toxicity in Mammals:
                             Implications for Risk Assessment
        Dallas M. Hyde, Robert P. Bolender*, Jack R. Harkemab and Charles G. Plopper
Department of Veterinary Anatomy and Cell Biology, School of Veterinary Medicine, University
of California, Davis, CA; Department of Biological Structure,3- School of Medicine, University of
Washington, Seattle, WA;  Inhalation Toxicology Research Institiute,^ Albuquerque, NM.

Abbreviated title: Morphometry and Risk Assessment
Address correspondence to:   Dr. Dallas M.. Hyde
                           Department of Veterinary Anatomy and Cell Biology
                           School of Veterinary Medicine
                           University of California
                           Davis, CA 95616

-------
                                    ABSTRACT
   Recent advances in quantitative morphology provide all the tools necessary to obtain structural
information in the lung that can be quantified and interpreted in the three dimensional world of
toxicology. Structural hierarchies of conducting airways and parenchyma of the lung provide 1)
numbers of cells per airway, lobe, or lung, 2)  surface areas of cells, airways, and alveoli,  3)
length of airways and vessels and 4) volumes of cells, alveoli, airways, vessels and individual
lobes or the entire lung. Unbiased sampling of these subcompartments of the lung requires
fractionation of lobes or individual airways. Individual airways of proximal and distal generations
are obtained by airway microdissection along one axial pathway and comparisons made between
airway generations.  Vertical sections of selected airways are used to sample epithelium and
interstitium.  Using this unbiased approach of quantitative morphology, we have shown that
 inhalation of low ambient concentrations of ozone (0.15 ppm) near or.at the United States National
 Ambient Air Quality Standard (NAAQS) (0.12 ppm ozone) induces significant alterations in
 bronchiolar epithelium and interstitium in nonhuman primates but not rats. The alterations do not
 appear to be concentrations time-dependent, thereby bringing into question the current NAAQS
 that may be at or above the threshold for distal airway injury in primates.

 Keywords: Morphometry, risk assessment, lung, design-based methods, stereology
                                             2 '

-------
                                  INTRODUCTION


     An unbiased morphometric assessment of pulmonary toxicity in animal lungs must consider


 the complexity and diversity of the entire branching airways and parenchyma. The extrapolation to


 humans of studies of toxic agents injurious to the respiratory system using animal models assumes
                                                            9

 comparability in the structure and function of the respiratory system of these model species and


 humans. The underlying assumption is that data, especially morphometric data of lung structure,


 obtained in model species can be extrapolated to humans.


     Ozone is the most reactive and toxic oxidant gas in photochemical air pollution. Ozone is also


 the principal air pollutant in many urban areas during the summer months.  In recent years,


 maximal monthly concentrations of ozone ranging from 0.1 to near 0.3 parts per million (ppm)


 have been reported in Mexico City. 1'2 Ozone inhalation produces injury in at least three target


 areas of the respiratory system of animals: nasal cavity, trachea and the centriacinar region.3-13


 Recent reports that some pulmonary function impairment was induced in exercising children and


 adult humans exposed to ozone concentrations at or near the ambient concentration in the  current


U. S. National Ambient Air Quality Standard (NAAQS) for ozone (0.12 ppm)  have opened to


question the health safety of the NAAQS.14'21


    The majority of studies defining the pathogenesis of ozone-induced injury have been


conducted using the laboratory rodents. Small laboratory mammals show an initial response of


cellular necrosis, exfoliation, degranulation of secretory cells, followed by epithelial hyperplasia


and hypertrophy.6.8,9,11,22,23  Experimental studies with rats at-concentrations near the


NAAQS, suggest that the lungs of rats are relatively insensitive to these environmentally relevant


concentrations of ozoneA24  Direct assessment of the human susceptibility to injury by ozone


inhalation is limited to physiologic assessment or measurement of bronchoalveolar lavage cells and


fluid following short-term, low level exposures;  thus, accurate assessment of the risk of ambient


ozone to humans is difficult.25,26


    Extrapolation from data collected using the laboratory rat suggests that there is minimal risk


for humans at ambient concentrations of ozone. In contrast, monkeys exposed to 0.15 ppm ozone

-------
for 6 or 90 days, 8 hours per day, had significant nasal epithelial lesions of ciliated cell necrosis,
attenuated cilia, and secretory cell hyperplasia and bronchiolar epithelial lesions of hyperplasia and
hypertrophy of nonciliated  bronchiolar epithelial cells and intraluminal accumulations of
macrophages.10,27 The bronchiolar lesion was also characterized by significant thickening of the
interstitium 27 These comparative quantitative results of differential sensitivity to ozone inhalation
of the nonhuman primate versus the rodent wer& only possible because of the use of carefully
applied  morphometric methods.   This paper  reviews  the state-of-the-art application of
morphometric methods to pulmonary tissue evaluated in the assessment of ozone-induced lung
injury.
                                      'METHODS
      We will use  three guiding principles in morphometry:  1)  Use  design-based methods to
 quantify structural changes,  2)  Use structural hierarchies to link and interpret experimental data,
                                                    
-------
 HIERARCHICAL DATA: The lung is a complex organ composed of numerous compartments
  ranging in size from molecules to tissues. Hierarchies allow us to organize data according to the
  size of the structures.  Hierarchy equations define the relationships among and link data within
  and across hierarchical levels.28  For example, if we desire the number of type I epithelial cells in
  the lung, we only need to multiply the number of type I epithelial cells per volume of interalveolar
  septal tissue by the volume density of the interalveolar septal tissue per parenchymal tissue, the
  volume density of parenchymal tissue per lung, and the volume of the lung. In this example the
  object at one magnification becomes the reference at the next higher level of magnification. Also,
  information is provided at each level or magnification and not just at the organ level. Two general
  guidelines emerge from hierarchy organization:  1) Use the lowest reasonable magnification
  (acceptable resolution) to increase sample size for measurements and 2) If major compartments
  and their subcompartments cannot be measured at the same magnification, then the magnification
  should be increased to optimize resolution in the subcompartmenL
 CRITICAL DATA:  A critical data set is required to detect and interpret quantitative data in any
 organ.28 These data include the volume of the structure, the number of cells in the structure, and
 the structural components or densities. The  critical data set allows  one to move data about a
 structural hierarchy, detect and interpret structural changes, and create links to other data types.
 ABBREVIATIONS, SYMBOLS, AND TERMINOLOGY:  Pulmonary   structures   can   be
 described as having volumes, surfaces, lengths, and numbers:  V, S, L, N.  Structural densities
 relate these parameters to a unit of reference volume: V/V, S/V, L/V, N/V (represented as Vy, Sy,
 LV, Ny). These four defining parameters are further defined by accompanying symbols. The
 symbol (i/ref) defines the ratio of two compartments as given for the densities:  Vv(i/ref),
 Sv(i/ref), Lv(i/ref), NyO/ref). The compartment of interest, the small'"!", is related to  the
reference compartment, "ref'.  For example, the volume density of capillaries within the lung is
represented as Vy(ca, lu). We  use the first two the first two letters of the compartment to
abbreviate the names. For example, lung becomes lu, trachea tr, capillaries ca and collagen co.
Extra letters can be added to avoid duplicates.

-------
TEST GRID SYSTEMS:  Point and intersection counting, using coherent test systems are more
efficient for collecting raw data  than digitizers requiring hand tracing of objects.31 Hence, for aU
stereologic measurements other than those that can be done by image analysis^ point and
intersection counting using test grid systems prevail.  It should be noted that even when image
analysis can be applied to lung tissue, such as quantising the volume of stored mucosubstance per'
surface area of epithelial  basal lamina, it is only 12-fold more efficient than manual methods.33
We prefer using the efficient point and intersection counting methods by employing coherent test
systems designed to give reference points (Pr ), reference line lengths (L, ) and reference areas
(Ar) after Weibel,31 CruZ-Orive34 and Baddeley et al,^  The counting rule of Gundersen36
 should be followed when making profile counts on any of the  grids.  The rule is to count all
 profiles totally within the counting frame that do not intersect the "forbidden lines" at any point.
 The forbidden lines include two adjacent borders of the frame (left and bottom as marked by solid
 lines on the grids) and extended lines from left top and lower right corners.
 VOLUME OF THE STRUCTURE: One of the most common starting points and our first critical
 data is the volume of the lung or its individual lobes. One of the most direct  methods  is to
 systematically cut fixed lung lobes into slabs of equal thickness, dehydrate, embed and section
 sampled slabs and determine all data within a volume that is common to all levels  of observation
 (e.g., a volume that is fixed, dehydrated, embedded and sectioned).  By incorporating a common
 reference into the experimental study, it is then possible to move data freely across the various
  levels of the hierarchy without fear of experimental bias. To estimate the volume of a lung lobe,
  take about ten samples from the slabs (selected systematically), determine their cumulative area by
  point counting, and multiple by the average slab thickness (Cavalieri method^.  The volume of the
  individual slabs can be estimated more precisely by defining their shape as a prismatoid using.
  computer digitization and analysis, but this level of precision  is unwarranted and not a marked
  improvement on the Cavalieri method of volume estimation .37
       Another approach, the optical volume fractional (OVF), provides estimates for the volume
   of the structure, the total number of cells in the structure, and the numerical density of cells.38 It

-------
 combines two of the primary tools of stereology, the fractionator39 and the optical "disector".40
 The fractionator systematically subdivides a structure into smaller and smaller fractions until a final
 fraction is obtained.  Volume is estimated in the final fraction by the Cavalieri method and related
 by fractions to estimate the volume of the entire structure.  If we count the number of cells in the
 final fraction using the OVF method, then we can estimate the number of cells per unit volume of
 compartment and within the entire structure.  The OVF method allows us to build structural
 hierarchies for the lung by establishing links between light and electron microscopy.  As long as
 specimens are treated the same (similarly fixed, dehydrated, and embedded) the links between light
 and electron microscopy are valid.
 NUMBER OF CELLS IN A STRUCTURE:  The second critical data required is the number of
 cells  in a structure (lung lobe, airway, vessel, alveolus, etc.). The most direct and unbiased
 method is to count cells in 3-D space.41.42  This is the basis of the "disector" principle. Counting
 methods based on the disector include the fractionator,40 optical fractionator,43  optical' volume
 fractionator,38 nucleator,44 and selector.45  The point sampled intercept,46 mean boundary,47
 and boundary sampled intercept48 methods use single random sections and count objects in 2-D
 space. They are largely unbiased for shape, but not for size.
    Using direct counting  of cells in 3-D space we are given three options for estimating the
 number of ceUs in the structure. If we want to estimate only total cell number, the fractionator or
 optical fractionator will be the easiest. Both methods are efficient and independent of shrinkage
 and swelling artifacts. For hierarchical studies, wherein  data are collected from several levels
 within the structure, we will want numerical density estimates for cells (number in the volume of
the various  compartments).  These estimates become critical for detecting changes in cell
compartments, such as organelles, because they allow us to calculate average cell data from
stereological densities. For example, assume we estimate cell numerical density of type n cells
within the volume of interalveolar septa (Ny(#2,is)) using the optical dissector and the volume
density of type H cells within the volume of interalveolar septa (Vv(ii4s), then we can calculate the
average volume of type II cells (V(ii)) by dividing the volume density  by the numerical density as

-------
follows:
V(ii,#2) = Vy(iUs) / Nv(#2,is) ,
where #2 represents the number of type E cells which in the denominator of V(ii,#2) becomes 1.
The real units for volume are in all in the same cm3 units and the reference volumes for VV and
NV are the same and thereby divide to 1.
OPTICAL DISECTOR:  The optical disector method counts cells directly in a measurable
volume 40,49   whether light or laser confocal microscopes are used to optically focus through a
thick section (usually about 20  ^m thick), a short depth of focus (1  to 3 ^m) is essential to
optically section the tissue and a length gauge required to precisely move in the Z direction.
 Usually a lens with a high numerical aperture satisfies the short depth of focus problem. This
 unbiased counting method is direct provide we use a 2-D unbiased counting frame36 and extend
 the counting frame concept by excluding structures counted on either the top or bottom of the cube.
 We estimate the reference volume by point counting an optical section in the middle of the cube that
 provides us with a reference area that is multiplied by the distance traveled in the Z direction for
 counting structures. Note we can use the area of the 2-D counting frame if it is totally filled with
 the reference area.
 DENSITIES:  A density is the ratio of two compartments, a compartment i in the numerator and a
 reference compartment, which is usually a volume in cm3, in the denominator. The four standard
  densities include volume, surface, length, and number.  Since the numerator and denominator in
  densities are both variables, they cannot reliably detect changes unless they are related to the
  volume of the structure.
       Two types of sections can be generated when structures are sampled systematically: (1)
  vertical sections (blocks rotated randomly about a vertical axis and sectioned), or (2) isotropic
  uniform random (IUR) sections (blocks oriented randomly in all directions and sectioned). Since
  the lung contains many aniostropic (oriented) structures, only systematic sampling and vertical
  sectioning guarantee unbiased estimates for all four densities.  Estimates of surface and length
   densities in the lung require a cycloid test grid oriented with respect to the vertical axis. Volume

-------
 and numerical densities can be made with vertical or IUR sections and thus the grid type is not
 specific.
 VOLUME DENSITY:  Volume density, Vy , is independent of the sectioning angle and the
 orientation of anisotropic structures because it affects both the object and reference phase equally.
 Volume density should be estimated by point counting techniques.  Point counting has been shown
 to be the most efficient method of estimation^ 1 and it uses the formula
 Vy(i,r) =  Pi/Pr
 where Pj is the number of point "hits" on the compartment of interest and Pr is the number of
 point "hits" on the reference compartment.50
 SURFACE DENSITY:  Surface density, Sy , is influenced by both the sectioning angle and the
 shape of anisotropic structures.^ 1  For isotropic structures, surface density can be defined .
 sVi,r =  2Ii/ LT
 where Ij is the number of intersections of the object surface and Lj. is the test system length of the
 reference component.52,53  xMs equation is valid for test lines that are isotropic uniform random
 in three dimensional space.  To meet this requirement using a lattice grid, the microstructures must
 be distributed uniformly and randomly and their orientation must be isotropic. The use of vertical
 sections, defined along the  plane of preferred orientation for anisotropic microstructures, and a
 cycloid test grid system,35  gives surface density estimates that correct for anisotropic orientation
 directly using the equation for Sy. Vertical sections alone, however, do not guarantee isotropic
 random encounters with the orthogonal test lines used to estimate surface and length densities in
 IUR sections. Sin weighted test lines alone the vertical axis, arranged continuously as cycloids
correct this bias of vertical sections.35  The requirements of vertical sections  according to
Baddeley et al.35 are as follows:  (a) Identify a vertical axis (along a preferred or arbitrary axis),
 (b) All vertical sections must be cut parallel to the vertical axis. The test grid must be oriented with
respect to the vertical axis, (c) All vertical sections must have random positions (systematic
 sampling of slices) and isotropic random orientation (spin about vertical axis), and (d) A test line
on vertical sections must be weighted proportional to sin 0, where 6 is the angle between the  test

-------
line and the vertical direction. Some examples of vertical sections are: (a) longitudinal sections of
skeletal muscle,  (b) sections of skin and other flat epithelial (e.g., tracheobronchial epithelium in
microscopic windows) normal to the exterior macroscopic surface, and (c) sections of an arbitrary
organ, obtained by cutting the organ into parallel slabs (with an arbitrary common direction) and
then placing some of the slabs on a flat surface (horizontal plane) and sectioning normal to the flat
surface. To obtain vertical sections of tubular organs (e.g., tracheobronchial airways), the organ
or airway must be opened along its axis, flattened along its abluminal surface that becomes the
horizontal plane. Relative to this defined horizontal plane, the vertical sections must be selected in
a random orientation. For test grid systems, superimposed on vertical sections, a test line is given
 a weight proportional to sin 6, where 9 is the angle between the test line and the vertical direction.
 Either a numerical weight for each intersection count obtained with test lines at a given  angle or a
 test system in which test lines at an angle 9 to the  vertical have length proportional to sin 9 is
 required for a correct weighting factor. The cycloid grid has a unique property where the tangent
 part of the curve is at an angle 9 to the vertical axis that has length proportional to sin 9.35 Thus,
 the sin 9 weighting factor is incorporated into the grid and with the intersection count per unit
 length of cycloid test curve gives  an unbiased estimate of surface density.
 LENGTH DENSITY:  Estimating lengths with IUR sections can become unusually problematic
  when linear structures have anisotropic orientations in tissues.54 A new design-based  method
  avoids this problem of anisotropy by using vertical sections and projected images.54,55 with
  vertical sections, all linear structures contained within the volume of a thick section or slice  are
-  projected onto a plane.  Counting  the intersections between a cycloid test line and the linear
   structure and measure the section thickness with a length gauge provides an unbiased estimate of
   LV.  To collect data with this method, a cycloid test grid must be oriented with respect to the
   vertical axis of the section, not perpendicular to it as with SV estimates. To estimate the length of
   capillaries per volume of interalveolar septa (LV(ca,is)) we need to count the total points hitting on
   the  reference component (is) and the number of intersections I(ca) between test lines and the
   capillaries. We then evaluate the Gokhale equation:54
                                              10

-------
Lv(ca,is) = 21 (t * 21(ca) /1 L(is)),
where t is the mean section thickness and X L(is) the total test line length in the reference
component (is).
NUMERICAL DENSITY: Numerical density, Nv , was introduced previously in the number of
cells in a structure. When NV is desired, one simply uses the optical disector and measures the
volume in which the cells are contained.  If the optical director is combined with the fractionator
method, it provides the most powerful method, the optical volume fractionator.38
STATISTICAL  CONSIDERATIONS:   Statistical considerations  relative to stereological
measurements have been presented in detail elsewhere.56-60 However, a compelling argument on
the contribution to the total variance of a group of animals in a stereological study of the harmonic
mean thickness of the glomerular basement membrane was provided by Gundersen and Osterby.^8
They showed that animals contributed 70%, blocks 19%, fields 8% and intercepts and measuring
3% of the total variance.  In our laboratory and those of others,61  the results of stereological
measurements of lung tissue agree with those in kidney and identify animal and block variance as
the greatest source of variation in a study.  A logical approach to this problem means that we
always use sufficient number of animals and blocks per animal in our studies. Then in turn we use
the minimal number of fields per block and points per field to estimate the block values. Using
ratio estimators,  we  sum the values over blocks and then use the means of the block values to
estimate the organ or animal value.56
FIXATION: Fixation and sampling are critical aspects of any morphometric study of the lung. If
the pulmonary epithelium is to be examined only by light microscopy, a 10% buffered formalin
solution is adequate.  However, the use of electron microscopy usually requires a glutaraldehyde-
paraformaldehyde fixation (440 mOsm, pH 7.4).62 since no true dimensions of cells and their
organelles are known, some investigators have employed quick-freezing and cryosubstitution to
compare morphological features with those in fixed tissues,63 or to examine unfixed antigenic
determinants. The relative osmotic effects of glutaraldehyde and buffer solutions were evaluated
on the. harmonic mean thickness of the blood/air barrier and the shape of erythrocytes in pulmonary
                                          11

-------
capillaries by instiUing fixative in lungs of rats.64 Mathieu et aL**> showed that 300 mOsm was
required to maintain eiythrocyte shape and that glutaraldehyde concentration exerted an osmotic
effect, even though it was less important than the buffer. In essence, the harmonic mean thickness
of the blood/air barrier varies inversely proportionally to the total osmolarity.  Our most commonly
used fixative is 1% glutaraldehyde, 1% paraformaldehyde at pH 7.4 and adjusted to 360 mOsm
with cacodylate buffer.  However, the goal of a study also dictates the fixative to be used.  If.
preservation of antigenic determinants is critical, then mild fixation may be required. This had to
be determined empirically in most cases. One percent paraformaldehyde adjusted to 360 mOsm
with cacodylate buffer and applied for about 30 minutes maintains the majority of antigens we are
interested in lung tissue.  Perhaps the most critical element of a morphometric study is that the
composition of the fixative, dehydrating and embedding solutions be constant.  This consideration
 should extend to embedding and sectioning tissue from all experimental groups at the same time.
 AIRWAY MICRODISSECTION: The approach we have used for precise sample selection is to
 employ airway microdissection with a specific binary numbering system.65  This can be applied to
 lungs preserved by a variety of methods. Airway microdissection has been used in lungs inflated
 at a  standard air pressure and fixed by perfusion of the pulmonary vascular bed with aldehyde
 fixatives 66 We have also employed specimens that have been frozen after expansion with air to
 TLC and lyophilized in a freeze-drying  chamber.  Beginning at the lobar bronchus,  the
 intrapulmonary airways and accompanying parenchyma are split down the long axis of the largest
 daughter branch or down the axial pathway of the primary airway. An attempt is made to expose
  as many minor daughter side branches as possible.  Each airway is numbered by a binary system
  originally described by Phalen el aL&  The binary system is a simple numbering system that
  provides a branching history for each segment of the conducting airways. Each time an airway
  branches, the larger of the two daughter branches, or the one with the smallest angle of deviation
  from tHe long axis, is designated with the number one. The smaller branch, or the one with the
  larger angle of deviation from the previous pathway, is designated with a 0. This number provides
  a branching history, the number of generations of branching, and a general size relationship of
                                            12

-------
' specific levels. From these dissected specimens, samples can be taken from specific locations with
 known generation number.
 TRACHEOBRONCfflAL AND CENTRIACINAR AIRWAYS
        The tracheobronchiaJ airways are uniquely characterized by ciliated pseudostratified
 columnar epithelium, a submucosa with glands and cartilage. Since epithelial and interstitial cell
 populations vary by airway and show different injury and repair patterns by airway generation,^
 it seems only appropriate to focus our attention on airways to individual generations.  Airway
 microdissection is the only practical way to sample airways and still retain knowledge of the
 generation number.65
        We will use an example (Fig. 1) of airway microdissection, fractionator, optical disector
 and local vertical section method to illustrate how one can estimate, the number of epithelial,
 interstitial or inflammatory cells for one airway generation. Airway microdissection uses airway
 casts in the appropriate species to select the best plane to select desired airways. Airways are
 exposed along the longest axial pathway with as many small branching airways as possible in one
 plane. A  binary system is used to uniquely record airway branching from the trachea.  Once an
 airway is selected, both halves are cut out of the lung for trimming and subsampling. The airway
 is cut transversely into 2 mm rings. Every third ring is selected using a random start, laid flat with
 the luminal surface up, rotated randomly and local vertical sections cut perpendicular to the
epithelial basal lamina (Fig. 1). Every fifth block is selected from the series of rings, and tissue
blocks are embedded and cut in alternating step serial sections (1 and 20 Jim). We use the OVF to
count cells in 20 Jim sections, and  1 \im sections to estimate Vv and Sv with local vertical sections.
With those unbiased measurements, we can calculate the volume of the airway wall and its
components, epithelium and interstitium; the number of epithelial and interstitial cells in the airway;
the total surface area of the airway and number of epithelial or interstitial cells per unit of surface.
Data and calculations related to Figure 1 are given below.
The volume of the fixed, dehydrated embedded airway wall (AW).
V(AW)  = f(D x f(2) x f(3) x SV(fi) = 3 x 5 x 900 x (2.5 X 10-7cm3) = 0.003375 cm3
                                          13

-------
where f(l)....f(3) are the fractions sampled at each level from airway rings (f(l)) to fields (f(3)).
The sums of the individual section/field volumes (V(fi)) are used to estimate the volume of the
airway wall. If we counted 100 epithelial cells in the sections with the optical disector, then we can
estimate the total number of epithelial cells (epce) in the wall of the airway as:
N(epce, AW) = «« x f(2> x f<3> * INce = 3 x 5 x 900 x 100 = 1,350,000
In turn, we can calculate the surface area of the epithelial basal lamina (epbl), a good reference
surface, of the airway indirectly from the Sv of epithelial basal lamina to the volume of the airway
wallas:
Sv(epbl,AW) = 2XI(epbl) / L(AW) = 650 cm2 / cm3
where L(AW) is the total length of the grid line in the reference space (airway wall). The total
surface of the epithelial basal lamina of the airway is simply
S(AW) -Svfcpbl. AW) xV(AW) = 650 cm2/cm3 * 0.003375 cm3 = 2. 1938 cm2
 In turn, the number of cells per surface of the airway is calculated as:
 NS(ep,AW) = N(epceAW) / S(AW) = 1350000 / 2.1938 cm2 - 6.1538 X itf / cm2
 One can also calculate the average volume of an airway epithelial cell as:
 V(ep,ce) = VV(epce,AW) / NV(#/AW) = 0.2498 / (4 X H* / cm? ) - 6.24 X lO'lO cm3 = 624 ^
 where
 Nv(WAW) = N(epce,AW) / V(AW) = 1350000 / 0.003375 cm3 = 4 X 10» / cm3
 With these calculations, one can detect hyperplasia, hypertropy and differential growth of airways
 under numerous experimental or disease conditions.
                                    DISCUSSION
        Ozone is well recognized as an oxidant injurant to the mammalian respiratory system.  A
  pertinent question, particularly considering the recent efforts to assess the costs of bringing major
  metropolitan areas in the United States into compliance with the current NAAQS for ozone at
  0.12ppm, is the level of risk posed  to human beings by exposure to current ambient levels of
  ozone. Most of the assessment of health risks and health effects of ozone have been based on
  extrapolation of findings reported using laboratory rats. Comparisons of quantitative measures of
                                            14

-------
  cellular changes occurring in 2 of the 3 target zones within the respiratory system demonstrate a
  substantial difference in the sensitivity of rats and nonhuman primates. The nonhuman primate
  appears to be at least 1 order of magnitude more sensitive at low-level concentrations of ozone than
  is the laboratory rat.
         The assumption that all species have the same susceptibility to injury from toxic substances
  such as ozone is one of the primary difficulties complicating attempts to extrapolate lexicological
  data from damage in rats to assessment of risks for human health. Clearly the respiratory system
  in various orders  of mammals is substantially different in architectural and cellular composition:
  including the nasal cavity, the tracheobronchial airways and the centriacinar region of the lung. All
  three of these regions are targets for oxidant injury from ozone. Recently quantitative information
  became available that  allowed direct comparisons, using the same types of morphometric
  measurements, of effects of inhalation of ozone on the respiratory system of species of different
  orders (Rodentia and Primates). These studies indicate that there are both topographic and cellular
  differences between the rat and macaque monkey.
        This review suggests that nonhuman primates have less ability to protect their respiratory
  system, at least from a cellular perspective, than do rats (i.e., their cells  are inherently more
  susceptible to ozone-induced injury than are those of the rat).  The nasal  cavity that could be
  expected to remove less ozone in the primate than in the rat is more affected at lower concentrations
  of ozone in the primate.  Similarly, the centriacinar region of primates is more susceptible to lower
  concentrations of ozone when compared with rats by an order of magnitude.  What these
  comparative findings imply for an assessment of risks to humans from inhalation of ambient
  concentrations of ozone is still questionable.  It is reasonable to conclude that basing the
  assessment of human  health risk on studies without  nonhuman  primates  and unbiased
  morphometric methods would at best grossly underestimate the potential susceptibility of humans
•  to chronic lung injury from ambient concentrations of ozone.  It should be noted that the new
  generation of designed-based, morphometric methods will allow more precise comparisons to be
  made in studies of ozone-induced injury in experimental animals and can now provide the essential
                                           15

-------
data required for a more accurate assessment of the risk from ozone inhalation on human health.
REFERENCES
1.     H. Bravo-Alvarez, G. Roy-Occtia, R. Sosa, R. Torres, "Tendencia del problema del la
       contaminacion atmosferica por ozono en la zona suroeste de la ciudad de  Mexico
       (tendency of the air pollution problem in south-west Mexico City)", Memoirs of fte'
       ^.ve.nth national Congress of Sanity and En-—-"Tl1 P"?ineering. (Oaxaca, Mexico,
       September 1990).
2.     L. Calderon-Garciduenas, A. Osorno-Velazquez, H. Bravo-Alvarez, Delgado^Chavez, R.
        Barrios-Marquez, "Histopathologic changes  of the nasal mucosa in southwest
        metropolitan Mexico City inhabitants", Am. J. Pathol. 140, 225-232 (1992).
 3.      L. W.  Schwartz, D. L. Dungworth, M. G. Mustafa, B. K. Tarkington,  W.  S. Tyler,
        "Pulmonary responses of rats to ambient levels of ozone: Effects of 7-day intermittent or
        continuous exposure",  Lab. Invest. 34,565-578 (1976).
 4.      C. G. Plopper, D. L. Dungworth, W. S. Tyler, C. K. Chow, "Pulmonary alterations in
        rats exposed to 0.2 and 0.1 ppm ozone: A correlated morphological and biochemical
        study", Arch. Environ. Health 34, 390-395 (1979).
 5.     G. A.  Boorman, L.  W. Schwartz, D. L. Dungworth, "Pulmonary effects of prolonged
        ozone insult in rats:  Morphometric evaluation of the central acinus", Lab Invest. 43,108-
         115 (1980).
 6.     W. L. Castleman, D.  L. Dungworth, L. W. Schwartz, W. S. Tyler, "Acute respiratory
         bronchiolitis: An ultrastructural and autoradiographic study of epithelial cell injury and
         renewal in rhesus monkeys exposed to ozone", Am. J. Pathol. 98, 811-840 (1980).
  T.     B. E.  Barry, F. J. Miller, J. D. Crapo,  "Effects of inhalation of 0.12 and 0.25 parts per
         million ozone on the proximal alveolar region of juvenile and adult rats", Lab Invest 53,
          692-704 (1985). •
                                            16

-------
8.     L. E. Fujinaka, D.  M. Hyde, C. G. Plopper, W. S. Tyler, D. L. Dungworth, L. O.
       Lollini, "Respiratory bronchiolitis following long-term ozone exposure  in bonnet
       monkeys: A morphometric study",  Exp. Lung Res. 8, 167-190 (1985).
9.     R. K. Moffatt, D. M. Hyde, C. G. Plopper, W. S. Tyler, L. F. Putney, "Ozone-induced
       adaptive and reactive cellular changes in respiratory bronchioles of bonnet monkeys",
       Exp. Lung Res. 12, 57-74 (1987).
10.    J. R.  Harkema, C. G. Plopper, D. M. Hyde, J. A. St. George, D. W.  Wilson, D. L.
       Dungworth, "Response of the Macaque nasal epithelium to ambient levels of ozone:  A
       morphologic and morphometric study of the transitional and respiratory epithelium", Am.
       J. Pathol. 128, 29-44 (1987).
                                  *
11.    B. C. Barr, D.  M. Hyde, C. G. Plopper, D. L. Dungworth, "Distal airway remodeling in
       rats chronically exposed to ozone",  Am. Rev. Respir. Dis. 137,924-938 (1988).
12.    W. S. Tyler, N. K. Tyler, J. A. Last, M. J. Gillespie, T. J. Barstow, "Comparison of
       daily  and seasonal exposures of young monkeys to ozone", Toxicology 50, 131-144
       (1988).
13.    J. R. Harkema, J.  A. Hotchkiss, R.  F. Henderson, "Effects of 0.12 and 0.80 ppm ozone
       on rat nasal and nasppharyngeal epithelial mucosubstances: Quantitative histochemistry",
       Toxicol. Pathol. 17, 525-535 (1989).
14.    M. Lippmann, P. Lioy, G. Leikauf, "Effects of ozone on the pulmonary function of
       children", Adv. Mod. Environ. Toxicol. 5, 423-446 (1983).
15.    P. J. Lioy, T.  A. Vollmuth, M. Lippmann, "Persistence of peak flow decrement in
       children  following ozone exposures exceeding the National Ambient Air Quality
       Standard", J. Air Pollut. Control Assoc. 35, 1068-1071 (1985).
16.    W. F. McDonnell,  R. S. Chapman, M. W. Leigh, G. L.  Strope, A; M. Collier,
       "Respiratory responses of vigorously exercising children to 0.12 ozone exposure", Am.
       Rev. Respir. Dis. 132, 875-879 (1985).
                                         17

-------
17.
18.
 19.
 20.
 21.
 22.
 23.
  24.
  25.
S. I. Gibbons, W. C. Adams, "Combined effects of ozone exposure and ambient heat on
exercising females" J. Appl. Physiol. 57, 450-456 (1984).
E. L. Avol, W. S. Linn, t.  G. Venet, D. A. Shamoo, J. D. Hackney,  "Comparative
respiratory effects of ozone and ambient oxidant pollution exposure during  heavy
exercise", J. Air. Pollut Control Assoc. 74, 804-809 (1984).
 F. J. Kulle, L. R. Sauder, J. R. Hebel, M. D. Chatham, "Ozone response relationships
in healthy nonsmokers", Am.' Rev. Respir. Dis. 132, 36-41 (1985).
H. Gong Jr., P. W. Bradley,  M. S. Simmons, D. P. Tashkin, "Impaired exercise
performance and pulmonary function in elite cyclists during low-level ozone exposure in a
 hot environment", Am. Rev. Respir. Dis. 134, 726-733 (1986).
 D. M. Spektor, M. Lippmann,  G. D. Thurston,  P. J. Lioy, J. Stecko, G. O'Connor, E.
 Garshick, F. E. Spelzer, C.  Hayes, "Effects of ambient ozone on respiratory function in
 healthy adults exercising outdoors",  Am. Rev. Respir. Dis. 138, 821-828 (1988).
 F. J. Stephens, M. F. Sloan, M. J. Evans, G. Freeman, "Early response of lung to low
 levels of ozone", Am. J. Pathol. 74, 31-58 (1974).
 C. G. Plopper! C. K. Chow, D. L.  Dungworth, M. Brummer, T. J. Nemeth, "Effect of
 low level of ozone on rat lungs. II.  Morphological responses during recovery and re-
  exposure", Exp. Mol. Pathol. 29, 400-41 1 (1979).
  C.  K. Chow, C. G. Plopper, M. Chiu, D.  L. Dungworth,  "Dietary vitamin E and
  pulmonary biochemical and morphological alterations of rats exposed to 0.1 ppm ozone",
  Envir. Res. 24, 315-324 (1981).
  U. S. Environmental Protection Agency.  Air Polity Criteria for r>7(W- and Other
               Oxidants.  Volume I, (U. S. Environmental Protection Agency, Research
          Triangle Park, 1986).
                                            18

-------
               28.
 26.     U. S. Environmental Protection Agency. Summary of Selected New Information "H
        Effects of Ozone on Health and Vegetation: Draft Supplement to Air Quality Cri^ri^ for
        Ozone  and Other Photochemical Oxidants. (U. S. Environmental Protection Agency,
        Research Triangle Park, 1988).
 27.     J. R. Harkema, C. G. Plopper, D. M. Hyde, J. A. St.  George, D. W. Wilson, D. L.
        Dungworth, "Response of macaque bronchiolar epithelium to 0.15 and 0.30 ppm ozone",
        Am. J. Pathol. (1993), in press.
        R. P. Bolender, D. M. Hyde, R. T. DeHoff, "Quantitative morphology of the lung: A
        new generation of tools and experiments for organ, tissue, cell, and molecular biology",
        J. Appl. Physiol.: Cell and Mol. Biol. (1993) in press.
        H. J.  G. Gundersen, E. B. Jensen, "The efficiency of systematic sampling in stereology
        and its prediction", J. Microsc. 147, 229-263'(1987).
        S. Ogbuihi and L. M. Cruz-Orive, "Estimating the total number of lymphatic valves in the
        infant lung with the fractionator", J. Microsc. 158,19-30 (1990).
        E. R. Weibel, Stereological Methods.  Vol I.  Practical Methods for Biological
        Morphometrv.  (Academic  Press, New York, 1979), pp. 3-7, 106-108, 139-150, 218-
        223, 322-331, 354-379.
32.     D. M.  Hyde,  D.  Orthoefer, D.  Dungworth,  W.  Tyler,  R. Carter,  H.  Lum,
        "Morphometric and morphologic evaluation of pulmonary lesions in beagle dogs
        chronically exposed to high  ambient levels of air pollutants", Lab. Invest. 38, 455-469
        (1978).
33.     J. G. Heidsiek, D. M.  Hyde, C.  G. Plopper, J.  A. St. George,  "Quantitative
        bistochemistry of mucosubstance  in tracheal epithelium of the macaque monkey", J.
        Histochem. Cytochem. 35, 435-442 (1987).
34.     L. M. Cruz-Orive,  "The use of quadrates and  test systems in stereology,  including
        magnification corrections", J. Microsc. 125, 89-102 (1982).
               29.
               30.
               31.
_
                                                       19

-------
35'.
36.
37.
 38.
 39.
 40.
  41.
  42.
  43.
   44.
   45.
A. J. Baddeley, H. J. G. Gundersen, L. M. Cruz-Orive,   "Estimation of surface area
from vertical sections",  J. Microsc. 142,259-276 (1986).
H. J.  G. Gundersen, "Notes on the estimation of the numerical density of arbitrary
profiles: the edge effect",  J. Microsc. Ill, 219-223 (1977).
D.  M.  Hyde, D.  J. Magliano, E. Reus, N.  K. Tyler,  S.  Nichols,  W. S. Tyler,
"Computer-assisted morphometry:  Point, intersection, and profile counting and three-
dimensional reconstruction", Micros. Res. Tech. 21, 262-270 (1992).
R. P. Bolender, J. Charleston, "Software for counting cells and estimating structural
 volumes with the optical volume fractional", Micros. Res. Tech. (1993), in press.
 H. J. G. Gundersen, "Stereology of arbitrary particles. A review of unbiased number and
 size  estimators and the presentation of  some new ones,  in memory  of William  R.
 Thompson", J. Microsc.  143, 3-45 (1986).
 H. J. G. Gundersen, P. Bagger, T. F. Bendtsen, S. M. Evans, L. Korbo, N. Marcussen,
 A. Moller, K. Nielsen, IR.  Nyengaard,  B. Pakkenberg, F. B. Sorensen, A. Vesterby,
 M. J.  West, "The new  stereological tools: disector, fractionator and point sampled
 intercepts and their use in pathological research and diagnosis", Acta. Pathol. Microbiol.
 Immunol. Scand. 96, 857-881 (1988).
 R. T. DeHoff, "Quantitative  serial sectioning analysis: Preview", J. Microsc. 131, 259-
  263 (1983).
  D. C. Sterio,  "The unbiased estimation of number and sizes of arbitrary particles using
  the disector", J. Microsc. 134,127-136 (1984).
  M. J. West, L. Slomianka, H. J. G. Gundersen, "Unbiased stereological estimation of the
  total number  of neurons in the subdivisions of rat hippocampus using the optical
.  fractionator", AnatRec. 231,482-492 (1991).
  H. J. G. Gundersen,  "The nucleator", J. Microsc. 151, 3-21 (1988).
   L. M. Cruz-drive,  "Particle number can be estimated using a dissector of unknown
   thickness:  The Selector", J. Microsc. 145,121-142 (1987).
                                             20

-------
46.     H. J. G. Gundersen, E. B. Jensen, "Stereological estimation of the volume-weighted
        mean volume of arbitrary particles observed on random sections", J. Microsc. 138, 127-
        142 (1985).
47.     F. Gittes, R. P. Bolender, "Counting cell nuclei with random sections: the effect of-shape
        and size", Micron. Microsc. Acta 18, 59-70 (1987).
48.     F. Gittes, "Estimating mean particle volume and number from random sections by
        sampling profile boundaries", J. Microsc. 158, 1-18 (1990).
49.    'V. Howard,  S. Reid, A. J. Baddeley, A. Boyde,  "Unbiased  estimation of particle
        density in the tandem scanning reflected light microscope",  J. Microsc. 138, 203-212
        (1985).
50.     E. Thomson, "Quantitative microscopic analysis", J. Geol. 38, 193 (1930).
51.     E. R. Weibel, Stereological Methods. Vol. 2.  Theoretical Foundations.  (New York,
        Academic Press, 1980).
52.     C. S. Smith, L. Guttman,  "Measurement of internal boundaries in three-dimensional
        structures by random sectioning", Trans AIME 197, 81-111 (1953).
53.     S. L. Tomkeieff, "Linear interceps, areas and volumes",  Nature 155, 24 (1945).
54.     A. M. Gokhale, "Unbiased estimation of curve length in 3D using vertical slices", J.
        Microsc., 159, 133-141 (1990).
55.     A. M. Gokhale, "Estimation of length density Ly from vertical slices of unknown
        thickness", J. Microsc. 167, 1-8 (1992).
56.     L. M. Cruz-Orive, "Best linear unbiased estimators for stereology", Biometrics 36, 595-
        605(1980).
57.     H. Elias H, D. M. Hyde, A Guide to Practical  Stereology. (Basel,  S. Karger A.G.,
        1983) pp. 16-24, 57-82.
58.     H. J. G. Gundersen, R.Osterby, "Optimizing sampling efficiency of Stereological studies
        in biology, or "Do More Less Well!"", J. Microsc. 121, 65-73 (1981).
                                         21

-------
59.    W. Nicholson, "Estimation of linear properties of particle size distributions", Biometrika
       57, 273-294 (1970).
60.    W. Nicholson,  "Application of statistical methods in quantitative microscopy",  J.
       Microsc. 113,223-239(1978).
61.    L. Chang, R. R. Mercer, K. Pinkerton, J.  D. Crapo, "Quantifying lung structure.
       experimental design and biologic variation in various models of lung injury", Am. Rev.
        Respir. Dis. 143, 625-634 (1991).
62.     E. E. Schneeberger-Keeley, M. J. Kamovsky, "The ultrastructural basis of alveolar-
        capillary membrane permeability to peroxidase used as a tracer", J. Cell Biol. 37, 781-
        793 (1968).
 63.     D. Sandoz, G. Nicolas, M. Laine, "Two mucous cell types revisited after quick-freezing
        andcryosubstitution", Biol. Cell 54, 79-88 (1985).
 64.    O. Mathieu, H. Claassen, and E. R. Weibel, "Differential effect of glutaraldehyde and
        buffer osmolarity of cell dimensions: A study on lung tissue", J. Ultrastruc. Res. 63,20-
        34 (1978).
 65.    C. Gr Plopper, A. T. Mariassy, L. O. Lollini,  "Structure as  revealed by airway
        dissection: A comparison of mammalian lungs",  Amer. Rev. Respir. Dis. 128, S4-S7
         (1983).
 66.     J- Gil, E. R. Weibel,  "Extracellular lining of the bronchioles after perfusion fixation of rat
         lungs forelectron microscopy", Anat. Rec. 169,185-200 (1971).
  67.     R. F. Phalen, H. C. Yeh, G. M. Schum, O. G. Raabe,  "Application  of an idealized
         model to morphometrv of the mammalian tracheobronchial tree", Anat'Rec. 190,167-76
          (1978).
  68.   D. M. Hyde, W. C. Hubbard, V. Wong, R. Wu,  C. G.  Plopper, "Ozone-induced
        tracheobronchial epithelial injury: Relationship to granulocyte emigration in the lung", Am.
        J. Respir. CellMol. Biol. 6,481-497 (1991).
                                            22

-------
FIGURE LEGENDS
Figure 1.  An illustration of airway selection, sampling and sectioning for estimating the total
number of cells in the airway, the mean volume of the cells and the number per surface of the
airway. Airway microdissection is used to exposed airways along the longest axial pathway with
as many small branching airways as possible in one plane. An airway is selected using both sides
of the dissected lobe, and cut into 2 mm transverse rings. Every third ring (f = 3) is selected by
stratified sampling with a random start, laid flat with the luminal surface up, rotated randomly and
local vertical sections cut perpendicular to the epithelial basal lamina. Every fifth block (f = 5) is
selected from the series of rings, it is embedded and cut in alternating step serial sections (20  [im
and 1 fim) and with a random start every ninehundreth section/field is selected (f = 900) to. estimate
Ny using the OVF method to count cells in a known volume, and Sy and Vy using point and
intersection counting. The "f' is for fraction which represent the fraction used for sampling and to
estimate total values for the sampled airway. Note that Sy requires a cycloid grid and a local
vertical section.
ACKNOWLEDGMENTS:    This work was  supported  in part by NIH grants HL28978,
ES00628, and DRR00169.
FOOTNOTES
a.  Cavalieri method is named for the Italian mathematician Bonaventura Cavalieri (1598-1647),
who first proposed the method for estimating volume.
                                         23

-------
X
                   X,
                                      Vertical xa&
     20
      Nv
Sv
Vv

-------
                          Daniel S. Marsman D.V.M., Ph.D.
                 Pathology Branch, Environmental Toxicology Program
                       Division of Intramural Research, NIEHS

Dr. Marsman's work has centered on understanding the underlying mechanisms of chemical
carcinogenesis. Prior to his current position at NIEHS he served as a postdoctoral fellow
at CUT, investigating the  rodent  hepatocarcinogenicity of the peroxisome  proliferating
chemicals.  His published work has included characterization of the cell proliferative and
promotional effects of chemicals and the persistence and metastatic potential of chemically
induced hepatocellular adenomas and carcinomas.  A veterinary (D.V.M., Michigan State
University) pathologist (Ph.D., University of North Carolina at Chapel Hill), his training has
also included an externship  to Khartoum, Sudan, investigating  Onchocerciasis  ('River
Blindness').  Currently, he is an Expert  Pathologist for the Environmental Toxicology
Program, NIEHS.   In addition to his responsibilities to the  National Toxicology Program
(NTP), he has recently been named team leader for developing and implementing a research
strategy at  NIEHS characterizing the potential risk  of the peroxisome proliferators.

-------
APOPTOSIS AND CHEMICAL CARCINOGENESIS
Daniel S. Marsman and J. Carl Barrett,
NIEHS, P.O. Box 12233, Research Triangle Park, NC  27709
ABSTRACT
Long recognized as  a normal component  of  organogenesis during development,
apoptosis ('programmed cell death') has recently been implicated in alterations of cell
growth and differentiation.  Tissue homeostasis is normally maintained by a balance
between cell division and cell death, with apoptosis often functioning in complement to
cell growth. Thus, antithetical parallels in chemical carcinogenesis can be drawn between
 apoptosis and the proliferate events more commonly  addressed.  As enhanced cell
 replication may contribute to an increased frequency of initiation, apoptosis within a tissue
 may counteract chemical carcinogenesis through loss of mutated cells.  Many strong
 carcinogens act as tumor promoters, selectively expanding an initiated cell population
 advantageously over surrounding cells. Similarly, chemicals with a selective inhibition of
 apoptosis would offer a growth advantage, while chemicals causing 'selective' apoptosis
 within mutated cells  would be expected to have an anticarcinogenic effect.  Selective
 apoptosis, in concert with cell-specific cell replication, may explain the unique promoting
  effects of different'cardrwgens such  as the peroxisome  proliferating  chemicals,
  phenobarbtal, and TCDD. Cell turnover, both cell growth and cell death, is central to the
  underlying processes of chemically induced carcinogenesis in animals, and the relevance
  of these effects to man.

-------
 Apoptosis: systematic, gene-directed cell daath




       During the development of an organism considerable remodeling takes  place,



 involving not only cell proliferation and differentiation, but highly organized apoptosis, or



 programmed cell death. No where has this been more elegantly demonstrated than in



 the nematode Caenorhabclitis eleaans where, in route to the adult organism of 959 cells,



 precisely 131 cells systematically die (1,2).  In this animal model, the genetic control of



 apoptosis involves unique  sets of controlling genes.  Some apparently  function in



 negative roles,  controlled by gene products in which gain of function mutations prolong



 the life of the cell, and deletions result in cell  death.  The targets of these inhibitors are



 the activators of the orderly and systematic process of cell death (1). Thus, the apoptotic



 process is highly active, with  numerous molecular events, some of which  can be



 recognized histologically.                  ,                               .




       Apoptosis, as recognized in liver and  other tissues,  has been well described



 elsewhere (3,4). In hepatocytes, these stages histologically involve cytoplasmic basophilia



 and nuclear condensation, eosinophilia and cytoplasmic condensation, and finally nuclear



 and cellular fragmentation and  dissolution.  Typically, when taking place within a solid



 organ, apoptosis also characteristically  involves phagocytosis-by  neighboring  cells,



 however luminal sloughing  or 'shedding' also occurs in epithelial  organs such as the



 kidney (5).   Early  cytoplasmic changes are considered an indicator of  the active



involvement of specific gene products used in the orderly demise of the cell, however to



date few definitive nonmorphological markers exist.  While active transcription of a few



genes is occurring, the majority of cellular functions are  diminishing, consistent with the



observed nuclear and cytoplasmic condensation. Coinciding with nuclear condensation



is the  induction  of an endonuclease, cleaving chromatin at internucleosomal DNA linker



                                      -2-

-------
regions and generating a characteristic oligonucleosomal ladder (-185 base-pair units).
observable following agarose gel electrophoresis (6).  The appearance of cytoplasmic
eosinophllia/condensation coincides with  fibrin organization, arranged  concentrically
around the periphery to collapse the cell on itself.  Phagocytosis by neighboring cells
does occur,  with apoptotic cells potentially utilizing  cell-surface signals involving the
vitroneotin receptor (7,8). Cellular blebbing, fragmentation, and packaging of detergent-
insoluble structures  appears  to  involve a  tissue  transglutaminase, however the
significance and generality of this marker for apoptosis remain to established (9,10). Final
degradation of the apoptotic remnant or 'body'  occurs within phagocytic cells and
 disappearance of  the apoptotic body may  occur  as rapidly as a  few hours  (11).
 However, smaller  degradative  membrane products (recognizable as lipofuscin) may
 remain for a considerable  period of time (12).
       Several examples of cell loss or irreversible growth arrest occur in mammalian
 systems besides apoptosis, including terminal differentiation, senescence, and necrosis
 and will only be mentioned briefly here (9). While markedly deferent morphologically and
 relative to their induction  and genetic control, all  of these forms of cell loss effectively
  result in exciusion of the cell(s) from the replicating  population. Terminal differentiation
  is associated wfth the expression of a specialized function or product of a tissue. Cellular
  senescence, is tightly controlled.by genes tnat are actuated or whose functions become
  manned at the end of the l«e span of the cell. Defects in the function of these gene
  products allow cells to escape the route to senescence and are thought to be 'immortal'.
  Senescence may be one mechanism by which tumor suppressor genes operate (13,14).
   Necrosis, in contrast to apoptosis, is  not a orderly  cellular process but  rather the
   disorganized death of a cell.  Random dissolution of all cellular components, the release
                                         -3-

-------
 of cellular constituents, and often the induction of an overt inflammatory response are all
 characteristics of necrosis, which  are not typically associated with apoptosis  (15).
 Apoptosis, even when involving a considerable portion of an organ or tissue, is not
 associated with an inflammatory process (16), although apoptosis and necrosis may co-
 exist (17).
       Despite considerable evidence that apoptosis is tightly controlled, only a few genes
 have been definitively shown to be involved. The cellular signalling and genetic control
 of apoptosis have been recently reviewed elsewhere and will not be discussed here (9).
 Candidate  genes which have been  identified include bcl-2, and the  tumor suppressor
 gene P53. Signalling factors suggested have included cytosolic Ca2+, IL-2, iL-3, and CSF.
 Paradoxically,  factors such as cytosolic Ca2+, and induction of c-fos and c-myc immediate
 early genes are shared by both the cell death and cell proliferation signalling pathways.

 Initiated cells:  impact of non-specific apoptosis and enhanced cell replication.
      Carcinogenesis has been described in risk assessment paradigms as a multistage
phenomenon involving multiple, discrete genetic mutations.  More recently mathematical
descriptions of  this phenomenon  have  become  increasingly   biologically  based,
recognizing the potential contributions of both cell proliferation and cell loss (18). These
models have been constructed to describe cancers involving two heritable changes ('M1'
and 'M2'; Figure  1), recognizing however that while useful for this discussion this is likely
an over-simplification.
                                      -4.

-------
Fig. 1  Biologically Based Cancer Model.

       Enhanced cell replication has  been implicated as a risk factor in several cases of
chemically induced carcinogenesis (19,20). In these instances, it is stated or implicitly
implied that background errors in DNA replication result in 'spontaneous' initiation of a
cell. Theoretically, in cases where either the replicating cell population ('N'; Figure 1) is
significantly increased or the replication rate of this population ('RO') far exceeds baseline
levels, the normally  rare probability  of a heritably altered cell (T) to form and persist is
increased.  While this 'probability-mutagenesis' is  often discussed  in relation to cell
proliferation, it is often neglected that a complementary role may be played by cell loss.
Thus, inhibition of apoptosis may similarly increase mutational frequencies by increasing
 the- population at-risk (N  or I populations).  However, while decreased cell loss  may
 impact on carcinogenesis by  the scenario addressed above, apoptosis-inhibition may
 itself impact directly on mutational frequencies (M1  and/or M2), as discussed below.

  Inhibition of apoptosis is reversible.
        One important characteristic  of the  inhibition of  apoptosis by chemicals  is
  reversibility.   A synchronized wave of apoptosis occurs within the liver following the
                                         -5-

-------
 withdrawal of numerous hepatic mitogens/hepatocarcinogens, indicative of homeostatic
 reversal of 'apoptosis-inhibition' (16,21-24).  Induction of apoptosis following withdrawal
 of endogenous hormones or growth factors has similarly been demonstrated (9,25,26).
 This regression is considered a normal homeostatic mechanism but suggests that direct
 evidence for apoptosis-inhibition is not evident in retrospective examinations. Within the
 context of chemical carcinogenesis  this observation also  suggests that apoptosis-
 inhibition by chemicals may be untraceable by examination of the end product, the tumor.
Apoptosis and initiation.                                              •
      The blockade of ceil death becomes even more intriguing when one considers that
aborted cell death may itself lead to initiation of the carcinogenic process. If apoptosis-
resistance is due to blockade of a stage of apoptosis after DMA fragmentation has begun
(6), outcomes of genomic instability such as deletions, frame shift mutations, and genetic
recombinants may increase. .In addition, cells that are normally programmed to die, due
to age or accumulation of spontaneous genetic damage, may persist and even replicate
to form  a heritably altered progeny if apoptosis is inhibited. Thus, apoptosis-inhibited
cells may not only have a growth advantage over their neighboring cells, but possess an
error-prone mutator phenotype, hypothesized to  be a critical event in some forms of
chemical carcinogenesis (27).  As mentioned above, apoptosis-inhibition is often through
a reversible alteration of signal transduction.  In these instances chemicals could act as
indirect initiators, with no demonstrable form of DNA reactivity. A feature of many diverse
hepatocarcinogens is  indeed their lack of DNA reactivity;  these chemicals are not only
strong promoters (28-30), but are also carcinogenic in long term feeding studies in the
absence of an initiator (31-33).
                                      -6-

-------
'Prnmotional factors' in tha nontext of a single mutated.celL
      Following the mutation of a single cell, tumor promotional factors (R1 and D1;
Figure 1) become critical to the 'survival' of this initiated cell population (I).  Survival of a
single initiated cell, assuming that cell loss is comparable to the surrounding tissue, can
be described as D1-DO.   Clearly, apoptosis-inhibition  directed at the  initiated cell
(D1 
-------
subpopulations of altered cells with a selective growth advantage over surrounding cells.


Enhanced hepatocyte apoptosis has been observed in the livers of rats fed the strong
                                                                                 *

rodent hepatocarcinogen, Wy-14,643 (12). In cases such as this, resistant hepatocytes


would have an accelerated growth advantage, with resistant cells potentially recruited into


cell proliferation due to the surrounding cell loss.





Cell-specific apoptosis and lesion regression.


      In  addition to  regression  of  the  liver  following the removal  of a  mitogenic


carcinogen, rapid induction of apoptosis is also observed within proliferative hepatic


adenomas (22-24). A hopeful line of research in cancer chemotherapy is attempting to


exploit apoptosis directed at specific cell-types, to  decrease directly tumor size rather


than simply to inhibit further growth.  High selectivity is still highly desirable however  as


complete loss of the tumor will not likely result unless the tumor is very small (i.e. the size


of I population is near 1; Figure 1), or the efficacy is  very high (D1  is significantly greater


than R1). In peroxisome proliferator-treated rats, many of the common histologic markers


indicative of preneoplasia  are negative, with numbers of hepatic foci often less in treated


animals than  in control (36,37). While this may be simply due to selective expression of


certain phenotypes, the relative decrease of individual hepatic foci may actually represent


loss  of initiated cell populations.  However, unless apoptosis is directed at specific cells,


only small populations of initiated cells would likely be eliminated. Cell-directed specificity


is not unlikely, as many of the most active promoters  are now thought to exert all or some


of their effects through receptor mediated events (eg. dioxins, peroxisome proliferators)


(38,39). Where this selective promotional activity is uniquely specific to a subpopulation


of cells or is of sufficient strength (R1 significantly greater than D1 and/or DO significantly


                                       -8-

-------
greater than D1), promoters take on a very distinct pattern of phenotypic expression.
Apoptosis and tumor progression.
      A hallmark of tumor progression is the persistence and autonomous growth of a
tumor, despite the removal  of the inciting agent.  As mentioned above,  escape from
cellular senescence or apoptotic-inhibition may lead to unrestrained growth of cells or
immortality (9). In hepatocellular tumors induced by the peroxisome proliferating chemical
Wy-14,643, tumor progression is particularly evident. Despite the induction of numerous
 large hepatocellular adenomas by Wy-14,643, withdrawal of this carcinogen results in few
 persistent tumors (Figure 2)1.   In contrast, hepatocellular carcinomas induced  by 52
 weeks of treatment are no  longer dependent on Wy-14,643 for either growth  or
 metastasis.   Escape from senescence or  loss of apoptotic  control  may  be one
 characteristic of the induced hepatocellular carcinomas.
               Adenomas
                                                          Carcinomas
   100
                                               40
        12   37   "
         Wy-14,643
        •Continuous1
    37/»
.104 -104 -104
 Wy-14,643
  •Stopped*
                                                                               0.25
                                                                               0.00
                                                     Wy-14,643
                                                    •Continuous'
  Fig. 2  BtotogM Potent*, of Hepatoce.lu.ar
                                         -9-

-------
 Conclusion

       Human  health  assessments  of cancer  risk  associated  with exposure  to

 environmental  chemicals  are  best when  a full  understanding of the carcinogenic

 mechanism in animals and the relevance of this response to man are known.  To this

 end,  understanding  the  chemical  induction and inhibition of apoptosis becomes  a

 necessary component in  our  understanding  of the  carcinogenic  mechanism and

 extrapolation of this data to humans.
1 Marsman, D.S. and Popp.J.A. (1993) Unpublished observations.
REFERENCES
1.


2.

3.
4.
5.
6.
 EIlis.R.E., Yuan.J.Y. and  Horvitz.H.R. (1991)  Mechanisms and functions of cell
 death. Annu. Rev. Cell Biol. 7:663-698.

 Raff.M.C. (1992) Social controls on cell survival and cell death. Nature 356:397-400.

 Bursch.W.,  Taper,H.S.,  Lauer.B. and Schulte-Hermann.R. (1985) Quantitative
 histological and histochemical studies on the occurrence and stages of controlled
 cell death (apoptosis) during regression of rat liver hyperplasia. Virchows Arch.
 [Cell Pathol.] 50:153-166.

 Kerr,J.F.R.,  Wyllie,A.H. and Currie,A.R.  (1972)  Apoptosis: A basic biological
 phenomenon with wide-ranging implications in tissue kinetics.  Br. J. Cancer
 26:239-257.

^Ledda-Columbano.G.M., Columbano.A., Coni.P., Faa,G. and Pani.P. (1989) Cell
'deletion  by  apoptosis during regression of renal hyperplasia. Amer. J. Pathol.
 135:657-662.

 Arends.M.J., Morris.R.G. and WyllieAH.  (1990)  Apoptosis:  the role of  the
 endonuclease. Amer. J. Pathol. 136:593-608.
                                     -10-

-------
8
9
"•
 
-------
 19.
20.
21.
Grasso.P., Sharratt.M. and Cohen.AJ. (1991)  Role of persistent, non-genotoxic
tissue damage in rodent cancer and relevance to humans. Annu. Rev. Pharmacol
Toxicol., 31:253-287.

Ames.B.N.  and Gold.LS. (1990)  Chemical carcinogenesis:  too  many rodent
carcinogens. Proc. Natl. Acad. Sci. USA, 87:7772-7776.

Tomei.LD., Kanter.P.  and Wenner.C.E.  (1988) Inhibition of radiation-induced
apoptosis jn vitro by tumor promoters. Biochem. Biophys. Res. Comm. 155:324->
331.
22.   Bursch.W.,  Lauer.B., Timmermann-Trosiener.l.,  Barthel.G.,  Schuppler.J.  and
      Schulte-Hermann,R. (1984) Controlled death (apoptosis) of normal and putative
      preneoplastic  cells  in  rat  liver  following withdrawal  of tumor promoters.
      Carcinogenesis 5:453-458.

23.   Gerbracht.U., Bursch.W., Kraus.P., Putz.B., Reinacher.M., Timmermann-Trosiener.l.
      and Schulte-Hermann.R.  (1990) Effects  of hypolipidemic drugs nafenopin  and
      clofibrate on phenotypic expression and cell death (apoptosis) in altered foci of rat
      liver. Carcinogenesis 11:617-624.

24.   Schulte-Hermann.R., Timmermann-Trosiener,!., Barthel.G. and Bursch.W. (1990)
      DMA synthesis, apoptosis, and phenotypic expression as determinants of growth
      of altered  foci in rat liver during phenobarbital  promotion. Cancer Res. 50:5127-
      5135.

25.   Rotello.R.J., Hocker.M.B. and Gerschenson.LE. (1989) Biochemical evidence for
      programmed cell death in rabbit uterine epithelium. Am. J. Pathol. 134:491 -495.

26.   Azmi.T.I. and O'Shea.J.D. (1984) Mechanism of deletion of endothelial cells during
      regression of the corpus luteum. Lab. Invest. 51:206-217.

27.   Loeb.LA.   (1991)   Mutator  phenotypes  may  be  required  for  multistage
      carcinogenesis. Cancer Res. 51:3075-3079.

28.   Cattley, R. C., and Popp, J. A. (1989) Differences between the promoting activities
      of the peroxisome proliferator Wy-14,643 and  phenobarbital in rat liver. Cancer
      Res. 49:3246-3251.

29.   Xu,Y., Maronpot,R. and Pitot.H.C. (1990) Quantitative stereologic study of the
      effects of varying the time between initiation and promotion on four histochemical
      markers in rat liver during hepatocarcinogenesis. Carcinogenesis 11:267-272.

30.   Pitot,H.C.,  Goldsworthy.T.,  Campbell.HA  and PolandA  (1980) Quantitative
      evaluation ' of  the  promotion,  by   2,3,7,8-tetrachlorodibenzo-p-dioxin   of
      hepatocarcinogenesis from diethylnitrosamine.  Cancer Res.  40:3616-3620.
                                     -12-

-------
31
32.
33.
 34.
 35.
36.
37.
     M  «m«n  n  9  Cattlev R C  Conway, J. G., and Popp, J. A. Relationship of


     [4 chloro-6-(2,3-xylidino)-2-pyrimidinylthio]acet,c ac,d (Wy-14,643) in rats. Cancer
     Res., 48: 6739-6744, 1988.

     Rossi L Ravera M., Raped, G., and Santi, L. Long-term administration of DDT
     or phenobarbital-Na in Wistar rats. Int. J. Cancer, 19: 179-185, 1977.

     H. iff IE Salmon A G Hooper.N.K. and Zeise.L (1991) Long-term carcinogenesis
     studie'lfo?2T7^ach.o?odibenzo-p-dioxin and hexach.orodibenzo-p-d.ox.ns.
     Cell biol. Toxicol. 7:67-94.



      Res. 49:6985-6988.
                                     (1977) Rapid emergence of carcinogen-induced
                                         • • *-- -•--  sequential analysis of liver
       carcinogenesis. Am. J. Pathol. 88:595-618.


       foci within the rat liver. Carcinogenesis 5:41-46.



        proliferator Wy-1 4,643. Carcinogenesis 3: 1231 -123d.
 39.
        4946.
        lssemann.1. and Green, & (1990) Activation of a member of the , «jt*J hormone
        receptor superfamily by peroxisome prolrferators. Nature 347.645-650.
                                        -13-

-------
                            BIOGRAPHICAL SKETCH:  DR.  DAVID MATTIE

              Dr.  Mattie is  a research toxicologist  in the Hazard  Assessment
              Branch, toxicology Division, Occupational and Environmental Health
              Directorate,  Armstrong Laboratory,  Wright-Patterson AFB,  OH   His
              primary area  of research interest is in dermal toxicology and risk
              assessment.   He conducts  research to  quantitatively  assess  the
              dermal route  of exposure for Air Force chemicals and materials.  He
              is also examining species differences in skin penetration.  He has
              developed   methodology  for  determining  skin:air  and   stratum
              corneum:air partition coefficients  for PBPK models  with a  skin
              compartment.                                                  s»*.J-n

              Dr. Mattie received his undergraduate degree in biology from Quincy
              College in  Quincy,  IL in 1974.  His Master of Science Degree  in
              Biology was received from the University  of  Dayton in 1977    He
              obtained  his Doctor of Philosophy Degree  in Biology from  the
              University  of Dayton in 1983.   He  is certified as a Diplomate  of
              the American  Board of Toxicology.

              He  was  Co-Chairperson  for  the   1992  Toxicology  Conference
              Applications of Advances  in Toxicology to Risk Assessment,"  as
                                       C°nference P^eedings that appeared  in
_

-------
 SIGNIFICANCE OF THE DERMAL ROUTE OF EXPOSURE TO RISK ASSESSMENT





             D.R. Mattie, J.H. Grabau, J.N. McDougal
                             ABSTRACT
          SS                    3
as compared to inhalation or oral routes
                                   ao
                                                     o    !
for systemic

permeability uuusuaiii..   ,,^^  r	~~~~' •„!*







































 development and acquisition.

-------
 INTRODUCTION

 BACKGROUND

 The skin, constituting  10 percent of total human body weight,
 acts as the major interface between the carefully regulated
 internal environment of the body and the comparatively dry and
 potentially hostile external environment.  The skin primarily
 functions as a barrier.  This barrier may still permit entry of
 chemical substances into the-body.  The potential for
 occupational or accidental skin exposure to nonvolatile and
 volatile chemicals (both of which may penetrate the barrier of
 the skin) requires a better understanding of chemical absorption
 through the skin to adequately determine risks of such exposures.

 Personnel working in the occupational environment are potentially
 exposed to a multitude of chemicals.  Maintenance, repair and
 fueling operations expose workers to engine oils, lubricants,
 fuels, hydraulic fluids, paints and solvents.  All of these
 classes of compounds present the potential for dermal exposure.
 As high as 40% of all occupational illness is related to skin
 disease (Wheeler, 1992).  For some substances, cutaneous
 absorption plays an important role in overall exposure
 (Scansetti, Piolatto, and Rubino, 1988).  The dermal route of
 occupational exposure was found to be the major contributor to
 total polychlorinated biphenyls body burden of transformer
 maintenance and repair personnel (Lees, Corn, and Breysse, 1987).
 The absorbed body dosage of xylene from hand contact with solvent
 mixtures over a cumulative period as short as 15 minutes was
 greater than that from inhalation over a full shift in auto body
 repair shops (Daniell et al., 1992).  Even wearing gloves is not
 an absolute form of protection as permeation of chemicals through
 gloves has been shown to occur (Perkins and Knight, 1988).
 Absorption of chemicals through the skin appears to be of greater
 significance than previously reported from industry or
 epidemiological experience (Grandjean et al., 1988).

 The dermal route of exposure may not be as significant as the
 inhalation route, but it can contribute to total exposure.  For a
 highly soluble chemical such as dibromomethane, the body burden
 from dermal penetration compared to inhalation for a rat is
 approximately 6% (McDougal et al.,  1985).   If a respirator were
worn but the skin was unprotected,  exposure to a soluble chemical
vapor would still occur.  A method to compare dermal vapor
exposure to inhalation exposure at the same concentration has
been described as a ratio of input functions for the contribution
of each route of exposure, providing the permeability constant,
 surface area of skin exposed and'aveolar ventiliation rate are
known or can be determined (McDougal and Clewell, 1990).

Chemicals in the liquid state must also be considered as  many
chemicals exist as a neat liquid or dissolved in a liquid medium
such as water.   Concentrations of pure liquid are much greater
than in their vapor form.   This results in greater penetration

-------
                                                              the
barrier properties of the skin.

Tsuruta (1975) and others have Deported on Jhe percutaneous









                                                             '
accurate assessment  of  actual  exposure  levels.

















 cSSSS'for^ScuSneoue absoption in order to assess its

 overall potential risk.

 ESTIMATES OF DERMAL PENETRATION

 Various methods have been used to ^^^fSSiiSy^onstant
 chemical to penetrate  through t]?ens^n;hjhjb?i™y of a chemical








  coefficient for skin is known.








  as shown in the following equation for flux.

-------
                       Flux =
= kp C
where C is the concentration gradient of chemical in the skin
(g/cm , 1 is the skin thickness  (cm), k^ is the solubility or
partition coefficent of the chemical in skin  (unitless), D is the
diffusion coefficient (cirr/hr), and kp is the permeability
constant (cm/hr).

Physical and chemical properties of chemicals such a's solubility
are important descriptors of skin penetration (Grandjean et al.,
1988; Pershing, Lambert and Knutson, 1989; and Surber et al.
1990).  The partition coefficient (PC), a measure of the affinity
of a chemical for tissue, is the ratio of concentrations at
equilibrium between the tissue and an adjacent media, such as
air, water or other vehicle.  Various experimental methods have
been reported in the literature for determining partition.
coefficient values for skin.  One method uses the octanol/water
partition coefficient as a surrogate for partitioning between the
skin (octanol phase) and the environment or vehicle (water phase)
(Bronaugh and Congdon, 1984; and Kasting, Smith and Cooper,
1987).  Octanol/water PC values are typically determined by
shaking the test compound in a mixture containing equal parts of
water and octanol.  After sufficient time for equilibration to
occur, the ratio of the amount of test compound in each solvent
is determined {Bronaugh and Congdon, 1984).  Hawkins and
Reifenrath (1985) compared octanol/water PC values to the percent
of applied dose of pesticides and steroid hormones after exposure
in vitro through pig and human skin.  Kasting, Smith and Cooper
(1987) used octanol/water PC values in a mathematical model to
calculate the flux of chemicals across the skin.  Berner et al.
(1988) used octanol/water PC values to confirm skin permeation
rates for a series of chemicals prior to examining the
relationship between the pKa of these chemicals and acute skin
irritation.  Octanol/water PC values have been used to calculate
dermal flux for setting a skin notation guideline for a Threshold
Limit Value-Time Weighted Average (TLV-TWA), (Fiserova-Bergerova,
Pierce and Droz (1990).   Although the octanol/water partition
coefficient has been used extensively in estimating dermal
penetration,  it is an oversimplification of the process of
chemical interaction with the skin.  The octanol/water partition
coefficient assumes that skin is homogenous with respect to
octanol.

Surber et al. (1990) measured SC/water and SC/isopropyl myristate
PC values.   In their study, partition coefficients were
determined'as a function of equilibration time, initial
concentration of drug in the vehicle, delipidization of stratum
corneum, and source and preparation of stratum corneum.  The
partition coefficients were considered as predictors of
percutaneous  penetration for the purpose of conducting dermal

-------
risk assessments (Surber et al., 1990).
TIERED APPROACH






use, initial toxicity results, etc.











            the endpoints for a Phase II screen.
 SKIN: AIR PARTITION COEFFICIENTS

 Introduction

 Th
 was developed in order to measure skin: air PC values.

 Material and Methods

 CHEMICALS:  The following chemicals were used for the




 halothane  f rom Halocarbon Labs,  Inc.  (HackensacK, NJ ) ,
 isoflurane f rom Anaquest (Madison, WI).




  Purina Formula #5008) were available ad libitum.  Tne amoien

-------
         FIGURE 1.   TIERED APPROACH TO DERMAL RISK ASSESSMENT
 INITIAL
 EXPOSURE
 ASSESSMENT
STRUCTURE ACTIVITY RELATIONSHIPS (SAR)
                                          IN VITRO IRRITANCY
                           PHASE I SCREEN
     DERMAL EXPOSURE
                                         IN VIVO ACUTE
                                                      SENSITIZ.
                   IN VITRO
                                       PHASE II  SCREEN
                                                    IN  VITRO
              PHASE III SCREEN
    CHRONIC BIOASSAYS
      SKIN PAINTING
       IN VITRO GENOTOXIC
            SCREEN
                       INIT/PROMOTION
                           PHASE IV SCREEN
ACTUAL
EXPOSURE
ASSESSMENT
    DERMAL RISK
    ASSESSMENT
           INHALATION
    RISK ASSESSMENT
HAZARD
ASSESSMENT

-------
temperature was maintained at 22±2°C and light was regulated on a
12-hour-light/dark cycle (starting at 0600 hrs).

SKIN PREPARATION:  Dorsal skin of a rat was clipped using an
electric clipper immediately after euthanasia with carbon
dioxide?  AftSr collecting the skin, the hypoderims was removed,
the skin was cut into 1 by 0.5 cm strips with a razor blade  and
Supelco, Inc., Bellefonte, PA).
PARTITION COEFFICIENT DETERMINATION:  Sample vials containing
skin and their corresponding empty reference vo.als w^e incubated
for 10 minutes at  32°C using a vortex evaporator to warm the akin
        and vials.  After  reaching 32°C  the caps were brxe fly
aat0ppm.oneor cemia
bag yielded a concentration of 406 ppm in the sample
                                                         1
                                                          .     Qf




 wi?h a flfSSionization detector (Model 5890A, Hewlett Packard,
toe reference   nd  sample vials  in  a  set,  sample
compared versus corresponding reference vials using
equation modified  from Gargas et al.  (I9«y).
                                                           ^^^
                                                         followxng
  PC - (reference -ea^ts) ,vlal
                     (test area cts)( sample volume)





 with  41  ppm (0.1  mL from the  lOOOOppm air bag)  or  102  ppm (0.25
 ^Irom  ?hS ioOOOppm air bag) dibromomethane  to examine  the
 effects  of different concentrations  on the  skin partition
 coefficient.

-------
 After  developing the technique  for determining a skin:air PC
 value  using  dibromomethane,  the procedure  was  used for the other
 volatile  organic chemicals of interest.  The time to reach
 equilibration was determined for each  additional chemical.
 Between 16-24 samples were measured to determine skiniair
 partition coefficients  for these chemicals at  203 ppm (5000 ppm
 bag).  Four  additional  samples  were measured at 41 ppm (1000 ppm
 bag) to confirm that there was  no concentration effect.

 STATISTICAL  ANALYSES:   To determine the equilibration time  and
 compare the  effect of different concentrations,  PC data was
 compared  by  a one way analysis  of variance.  Linear regression
 and correlation for  figures  comparing  PC values  versus
 permeability constants  and octanol/water PC values were
 determined using the line fit procedure in RSI  on a mainframe
 computer  (BBN Software  Products Corporation, Cambridge, MA).  The
 level  of  significance was accepted at  p<0.05.

 Results

 After  incubating samples for various times, the  equilibration
 time for  dibromomethane in skin was determined  to be 4 hours.
 The skin:air partition  coefficent for  dibromomethane at 203 ppm
 was 68.3+3.1.   The skin:air  was also determined  for 102 and 41
 ppm.   There  was  no significant  difference  between the skin:air PC
 values determined at 203, 102,  and 41  ppm  concentrations  (data
 not shown).

 The skin:air PC  values  for selected volatile organic  chemicals
 are shown in Table 1.  Approximately 19 samples were  analyzed for
 each chemical.   The  most common equilibration time  for this  group
 of volatile  chemicals was 4  hours.  The skin:air  PC values  ranged
 from 1.9  for hexane  to  91.9  for styrene.   For each chemical  4
 samples were also  run at 41  ppm.   The  value for the  skin:air PC
 at 41 ppm was always the same as  the value at 203 ppm (data not
 shown)  .

 The octanol/water  PC values  (Leo,  Hansch,  and Elkins, 1971) were
 compared to  the  corresponding skin:air PC values  for  11 of the above
 chemicals.   There was no correlation (r2=0.09)  between the
 octanol/water PC values and  skin:air PC values  (Figure 2).

 Permeability constants were  available  for  9 of the chemicals for
which skin;air PC values were measured in this study  (McDougal et
 al, 1985;  McDougal et al, 1986;  and McDougal et al,  1990).  There
was good correlation  (r^=0.93)   between permeability constants and
 skin:air PC values (Figure 3).

 Discussion

The skin:air PC values were compared to both octanol/water PC values
and permeability constants.  Octanol/water PC values have

-------
Table 1.  SkinrAir Partition Coefficients for Selected Volatile
Organic Chemicals
CHEMICAL

perchloroethylene
trichloroethylene
benzene
hexane
toluene
xylene
styrene
methylene chloride
carbon tetrachloride
methyl chloroform
halothane
isoflurane
                        SKIN:AIR PC

                          ( + S.E.)

                          41.5+1.2
                          31.8+1.5
                          34.5+1.9
                          •  1.9+0.1
                          43.0+1.8
                          50.4+1.7
                          91.9+6.8
                           13.6+0.5
                           12.4+0.6
                           10.8+0.6
                           10.6+0.7
                            4.5+0.3
       EQUILIBRATION
           TIME
n        (hours)

16          4
19          4
19          4
18          4
16          4
24          2
20          3
17          2
24          4
18          4
17    •      3
16          6
been used as a qualitative -asure of slcin
and Congdon, 1984; Hawkms and       ra
and Cooper, 19.87; Anderson and
                                              and Tayar  t al.,
                                              a      ghowed a
 the solvent Pr°P;F^es°g9 >    The  data  in  this  study suggests
  (Anderson  and Raykar,  1989).   -^f  °*  indicator  of  the relative
 that  skinrair PC values  are  a better  ^dicator  o     rmini   a
  skin  permeability  for  these  v°J;a^^iai g^een  to  identify the
  permeability constants.

  PBPK MODELING
  chemicals into tissues.  A PB-PK »odel

                                                  SSSS-

-------
Fig.  2.  Comparison of  skin:air  partition  coefficients with
octanol/water PC  values.
 CD
  >
 0
1500
1400
1300
1200
1100
1000
 900
 800
 700
 600
 500
 400
 300
 200
 100
  0
                 T	1
                       T	r
                                        —i—i	1—r
                                         Xylene
                                                      -i	r
                            Perchloroethylene
      Carbon Tetrachloride
         €

   - Hexane
    i

        •
     Methyl Chloroform

    Methylene Chloride
       Toluene


Trichloroethyle'ne
• Benzene
                                              Dibromomethane
                                             J	'*i   i   '
           5  10 15 20  25  30 35 40 45 50 55 60  65  70 75 80 85 90 95
                            Partition Coefficients
Fig. 3.  Comparison of  skin:air partition coefficients  with
permeability constants.
   E
   CO
   •8
   
-------
            and elimination of a chemical are then mathematically
H«ed for each compartment which has such a process.  The skin: air
described for eacn co p      equation in the dermal compartment

                                                                  -


iituation by extrapolating across -P--|0-nc|nSratnxtrapo?Sat?on

dermal exposure  standards .
Previous work with  PB-PK models  in  this
demonstrated their  ujefulneBS  xn ertrapolatjon^nd  the  risk

            PSSS  ifS"~990,f Fi.£?7rt  al.,  1989;  Gargas  et
          ,  ppitz  et al.,  1988; Ramsey and Andersen, 1984).
  ki-i iartitioS  cSiwicients  for volatile chemicals  have  been



^nemattcal'description  that adequately describes  the
biologically  relevant data (McDougal, 1991).
 Whole thickness skin, used in the Bkin:air J.
 I,*  __-, iaver which may be important in determining the






 physiological compartments.
 epo     to benzene vapor,  xposure to
 cell on the dorsal skin, and exposure to

                              oinslan^foi
                                             benzene
                                                            cl osed
                        lofr^ J^T^S 5|SeoSf sS?u Jons was one
                                     11

-------
half the human permeability  constant value  used  for dermal  risk
assessment,  0.111  cm/h.  Rat skin has  been  reported to  be more
permeable than human  skin by a  factor  of two to  four (McDougal et
al., 1990),  so the rat permeability constant was expected to be
at least twice as  high as the human value for benzene.


SPECIES DIFFERENCES IN SKIN  PENETRATION

INTRODUCTION

In an attempt to better understand factors  affecting dermal
penetration  and to be able to better extrapolate between animal
species and  humans, a study  was inititated  to quantitate selected
anatomical differences in skin  from a  number of animal  species.
Anatomical differences which may affect permeability include
density and  size of hair follicles, density of sebaceous, and
apocri-ne glands, capillary density and distance from the surface,
as well as thickness df the  various layers  of the epidermis and
dermis.  Animals used in the laboratory usually have fur compared
to human skin.  The skin of  animals provides insulation to cold
while human  skin serves an evaporative cooling function.  The
degree of regional variation is low in animal species while
regional variation is very high for human skin.  The species and
strains to be tested for differences in skin penetration are
mouse - B6C3F1 and Hairless;  rat - Fuzzy and Fischer; guinea pig
- Hartley and Hairless; Swine - farm;  and Monkey - Rhesus.

Permeability constants (cm/hr) will be determined for each of
three model chemicals in species which show the widest
differences in structure.  The chemicals, 1,2 dichlorobenzene,
perfluoroheptane,  and chloropentafluorobenzene, were chosen on
the basis of their structural characteristics associated with'a
different class of potential toxic chemicals.

In order to understand the effect of skin structure  on
permeability, anatomical and physiological  differences between
the species will be correlated with the the permeability constant
for each of the three model chemicals.   This analysis will
distinguish which are the important parameters that  affect
permeability in skin as indicated by permeability constants.

Mathematical descriptions of the anatomical factors  found to
affect skin permeability will be used to entend physiologically-
based pharmacokinetic (PBPK) models.   The PB-PK models will
ultimately be used to extrapolate the toxic effects of various
chemicals between species.   Once the PBPK models for the above
chemicals and species are fully developed,  the models will be
validated by additional experimentation in another species.
Following model development and validation in animals, the
ability of these models to predict the consequence of human
dermal exposure to toxic chemicals will be tested by comparing
permeability data predicted by PBPK models  with human data.
                                   12

-------
METHODS


Scions of  dorsal
                                   paraffin   one sat


sections was stained wltV^» SSl^t walconducted on sections

with Massons trichrome.  ^^/"^f^?"" |2m  (Quantimet 570C,

from each strain using an ""a|J ^Jf JSuSd were thickness











the basement membrane of the  epidermis .
 RESULTS


 Preliminary data showed that the hairless guinea pig and farm^ig















 volume .



 DISCUSSION

 Mditionalskin sepias are being analy^d^to C0onfir,








 §44 rats and Hartley and  Hairless guinea pigs .
  SUMMARY



  SUrf SltSf iS S^SST ff
  skin and its significance in risk assessment
  ACKOWLEDGEMENTS
  assistance.


  The animals  used in this study were handled in accordance with
                                    13

-------
the principles  in  the  Guide for the Care and Use of Laboratory
Animals,  prepared  by the  Committee  on  Care and Use of Laboratory
'Animals of  the  Institute  of Laboratory Animals Resources,
National  Research  Council,  DHHS,  National Institute of Health
Publication #86-23,  1985, and  the Animal Welfare Act of 1966,  as
amended.
REFERENCES

Andersen, M.E., H.J. Clewell  III, M.L. Gargas, F.A.  Smith  &  R.H.
Reitz  (1987) Physiologically  Based Pharmacokinetics  and  the  Risk
Assessment Process for Methylene Chloride.  Toxicol.  Appl.
Pharmacol. 87,  185-205.

Anderson, B.D.  and Raykar, P.V  (1989) Solute structure-
permeability relationships in human stratum corneum.  J.  invest.
Dermatol. 93, 280-286. •

Berner, B., Wilson, D.R., Guy,  R.H., Mazzenga, G.C.,  Clark,  F.H.,
and Maibach, H.I. (1988) The  Relationship of pKa and Acute Skin
Irritation in Man. Pharmaceutical Res. 5, 660-663.

Bronaugh, R.L.  and Congdon, E.R. (1984) Percutaneous  absorption
of Hair dyes:   Correlation with partition coefficients.  J.
Invest. Dermatol. 83, 124-127.

Clewell III, H.J. and M.E. Andersen (1985)  Risk Assessment
Extrapolations  and Physiological Modeling.  Toxicol.  Ind. Health.
1, 111-131.

Clewell III, H.J. and M.E. Andersen (1989)  Improving  Toxicology
Testing Protocols using Computer Simulations. Toxicol. Lett. 49,
139-158.

Corley, R.A. , Mendrala, A.L., Smith, F.A.,  Staats, D.A., Gargas,
M.L., Conolly, R.b. , Andersen,  M.E. and Rietz, R.H.  (1990)
Development of a Physiologically Based Pharmacokinetic Model for
Chloroform. Tox. and Appl. Pharm. 103, 512-527.

Daniell, W., Stebbins, A., Kalman, D., 
-------
Sci. 63, 479-510.

        M T   ME  Andersen & H.J. Clewell III  (1986)  A



Pharmacol. 86, 341-352.




Toxicol. Appl. Pharmacol.  98,  87-99.
"Skin" Deotation.

Hawkins, 0.8. and ^
                                     Med.  14,
 Water.  JRisA: Analysis 10(4), 581-585
 Pharmacol. and S^rin 1, 138-153
 Hyg. Assoc. J. 48, 257-264.

          , J...
                  j on  r^i i TTT  H J
 McDougal, J.N. and Clewe H J"r H.J..




 313-317.
     l.  Pharmacol. 79, 150-158
  vapors in rats and humans. Fundam.
                                                              In
                                              .  75. 378-381
                                              Dermal to Inhalation
                                                             -to-
                                   15

-------
McDougal, J.N., Jepson, G.W., Clewell  III,  H.J.,  MacNaughton,
M.G. and Andersen, M.E., (1986) A physiological  phannacokinetic
model for dermal absorption of vapors  in the  rat.  Toxicol.  Appl.
Pharmacol. 85, 286-294,

Morgan, D.L., Cooper, S.W., Carlock, D.L..,  Sykora,  J.J.,  Sutton,
B., Mattie,  D.R., and McDougal, J.N. (1991) Dermal  Absorption of
Neat and Aqueous Volatile Organic Chemicals in  the  Fischer  344
Rat. Environ. Res. 55, 51-63.

Perkins, J.L. and Knight, V.B. (1989)  Risk Assessment of  Dermal
Exposure to  Polychloririated Biphenyls  Permeating  a  Polyvinyl
Chloride Glove. Am. Ind. Hyg. Assoc. J. 50, A-171-172.

Pershing, L.K., Lambert, L.D., and Knutson, K (1989) Partition
Coefficient  and Solubilities of Estradiol in  a  Variety of
Vehicles .Predict the In Vivo Flux Across the  Human  Skin Sandwich
Flap. Clin.  Res. 37, 727A.

Ramsey J.C.  and Andersen, M.E.  (1984)  A Physiologically Based
Description  of the Inhalation Pharmacokinetics  of Styrene in Rats
and Humans.  Toxicol. Appl. Pharmacol.  73, 159-175.

Reitz, H.R., J.N. McDougal, M.W. Himmelstein, R.J.  Nolan, and
A.M. Schumann (1988) Physiologically-Based Pharmacokinetic  Modeling with
Methylchloroform: Implications for Interspecies,  High Dose/Low
Dose, and Dose Route Extrapolations. Toxicol. Appl. Pharmacol.
95, 185-199.

Sato, A. and Nakajima, T. (1979) Partition Coefficients of  some
Aromatic Hydrocarbons and Ketones in Water, Blood,  and Oil. Brit.
J. Indust. Med. 36, 231-234.

Scansetti, G., Piolatto, G., and Rubino, G.F. (1988) Skin
Notation in  the Context of Workplace Exposure Standards. Amer. J.
Indust. Med. 14, 725-732.

Surber, C., Wilhelm, K.-P., Maibach, H.I., Hall, L.L., and  Guy,
R.H. 1990) Partitioning of Chemicals into Human Stratum Corneum:
Implications for Risk Assessment Following Dermal Exposure. Fund.
Appl. Tox. 15, 99-107.

Tayar, N.E., Tsai, R.-S., Testa, B., Carrupt, P.-A., Hansch, C.,
and Leo, A.  (1991) Percutaneous Penetration of  Drugs: A
Quantitative Structure-Permeability Relationship Study. J.
Pharmaceutical Sci. 80, 744-749.

Tsuruta, H.  (1975) Percutaneous Absorption of Organic Solvents.
1. Comparative Study of the In Vivo Percutaneous Absorption of
Chlorinated  Solvents in Mice. Indust.  Health  13, 227-236.

Wheeler, K.F. (1992) Barrier Lotions, Along With Gloves, Can Help
Deter Occupational Dermatitis.  Occupational  Health and Safety,
Jan., 60-61.
                                  16

-------
BIOSKETCH

Thomas M. Monticello, D.V.M., Ph.D.

Dr. Monticello received his D.V.M. from Michigan State University
and following two years in clinical practice, completed a pathology
residency  at  North  Carolina  State  University's  College  of
Veterinary Medicine, a Ph.D.  in pathology  from  Duke University
Medical center, and a post-doctoral training program in toxicology
and experimental  pathology  at the Chemical Industry Institute of
Toxicology  (CUT)  in Research Triangle Park, NC.    Dr. ^Monticello
has extensive experience in upper respiratory tract toxicology and
pathology  and has authored numerous  publications  in this  field.
Specific  areas  of  interest  include pathogenesis  of chemically
induced  diseases  and neoplasia  of  the  respiratory tract,  cell
kinetics   and  carcinogenesis,   and  comparative  medicine  and
             A Diplomate of  the American  College of Veterinary
               he  is currently  a   Senior  Pathologist  in the
                Experimental  Pathology  at  Bristol-Myers  Squibb
pathology.
Pathologists,
Department  of
U*^* KSU .&. WU1S— A A W  ^^ J-  *rf*fcjj^ ^^^. .»-»••—•-- —- — —  —      -r ~.
Pharmaceutical  Research Institute in Princeton,  NJ.

-------
                         CELL PROLIFERATION AND
         FORMALDEHYDE-INDUCED RESPIRATORY CARCINOGENESIS
                   Thomas M. Monticello^ and Kevin T. Morgan^



     1 Bristol-Myers Squibb Pharmaceutical Research Institute, Princeton, NJ, and



                       2CHT, Research Triangle Park, NC
Running title: Cell Proliferation and Formaldehyde Cancer

-------
Abstract

       Formaldehyde is a nasal carcinogen in the rat but the cancer risk this chemical
poses for humans remains  to be  determined.   Formaldehyde  induces  nonlinear,
concentration-dependent increases in  nasal-  epithelial cell proliferation and  DNA-
protein cross-link formation following short-term exposure.  Presented in this review
are  results from a mechanistically-based formaldehyde inhalation study in which an
important endpoint was the measurement of cell proliferation indices in target sites for
nasal tumor induction.  Male F344 rats were exposed to 0, 0.7,  2, 6, 10 or 15 PPm
 formaldehyde for up to two years  (6h/d, 5d/wk).   Statistically significant increases in
 cell proliferation were confined to the  10 and 15  ppm groups  and which remained
 elevated  throughout the  study. The  concentration-dependent increases  in  cell
 proliferation correlated strongly with  the  tumor  response curve, supporting  the
 proposal that sustained increases in cell proliferation are an important component of
 formaldehyde carcinogenesis.  The nonlinearity observed in formaldehyde-induced
 rodent nasal cancer is consistent with a high-concentration effect of regenerative cell
 proliferation of the target organ coupled with the genotoxic effects of formaldehyde.  Cell
 kinetic data from these studies provide important information that may be utilized in the
 assessment of risk for humans exposed to formaldehyde.

 Key Words: rat, formaldehyde, cell proliferation, nasal cancer, risk assessment

-------
 Cell Proliferation and Cancer

       The complex process of cancer develops secondary to one or more mutational
 events that alter growth regulatory genes of normal cells, with subsequent clonal growth
 of the resulting precancerous or cancerous cells (1-2).  Cell proliferation, an essential
 component of the multistage process of carcinogenesis, is required for both initiation and
 promotion of neoplasia in certain organs, and it plays an essential role in the later stages
 of carcinogenesis, including the progression of benign  lesions to  malignancy and
 metastasis (3-4).   Each time  a  cell  divides there is a chance,  albeit rare, that a
 mutational event related to the carcinogenic process will occur (6-7).  Enhanced cell
 proliferation may increase  the  frequency of  these  spontaneous mutations either by
 errors in replication or by the conversion of endogenous or exogenous DNA adducts to
 mutations before DNA  repair can occur (8).  Chemicals that induce cytotoxicity and
 sustained  increases in  cell proliferation, therefore, could  enhance  the likelihood of
 cancer development by providing additional cell  divisions, each with an opportunity for
 somatic or chemically-induced mutations (7, 9).
       Epidemiologic evidence indicates that increased cell proliferation  induced by
external or internal stimulation is a common denominator in the pathogenesis of many
human cancers.  Prolonged irritation by physical or chemicals agents may cause cell
death, and the subsequent cell division that occurs during repair of the damaged tissue
may eventually lead to a cancer at the irritated site (10).  For example, tobacco, an
established carcinogen, is an well known irritant.  Snuff users develop leukoplakia, and
eventually, cancer of the buccal mucosa at the site of snuff application. Tobacco smoke is
a local irritant to the  epithelial tissue lining  the bronchi, lungs,  larynx,  pharynx,  oral
cavity and esophagus, sites where smoking-related cancers arise.

-------
       Many chemicals identified as carcinogenic for humans are genotoxic and have also



been determined to induce cancer in laboratory animals (11).  The primary biological



activity of a genotoxic chemical or its  metabolite, is the  alteration  of the  genetic



information encoded in the DNA, inducing a mutation in growth regulatory genes (12).



A genotoxic chemical administered at a dose that is both cytotoxic and a  cell proliferation



enhancer, would be expected to be a more effective mutagen and carcinogen then when



given at a noncytotoxic dose which does not induce cell proliferation  (7-8).  In  addition



to  regenerative  cell  proliferation  and cytotoxicity, other  cellular responses  such as



metabolic activation and DNA repair could also greatly affect the carcinogenic response



of  a target tissue to a given dose of a  chemical (9).  Understanding  the  relationship



between chemically induced cell proliferation and carcinogenic activity would be of value



 in the investigation of mechanisms of carcinogenesis, the selection of appropriate doses



 for cancer bioassays, and the improvement of risk assessment models (13).







        Research with some respiratory carcinogens, such as formaldehyde  gas,



  illustrates the principle that  both genotoxicity and enhanced cell  proliferation of the



  target organ should be considered  in mechanistic studies and the improvement of risk



  assessment models.  An extensive database on cell proliferation and the  induction of



  upper respiratory tract tumors has been generated from a long-term, mechanistically-



  based, formaldehyde bathogenesis study in the F344 rat (14).   This review presents



  time-course and concentration-response  data on  site-specific  increases in  nasal



  epithelial cell proliferation, and compares these data with the formaldehyde-induced



  tumor response. The application of these data in improving the risk  assessment process




  will also be addressed.

-------
 Formaldehyde





       Formaldehyde  is  an important  commodity  chemical used  widely in the


 manufacture of resins, particle  board, plywood, textiles,  and many other consumer


 products.  The finding that formaldehyde is a nasal carcinogen in  rats (15-17) has


 provoked concern that this chemical may also pose a cancer risk for humans. Although


 there is widespread, exposure of humans to formaldehyde, epidemiological data for


 exposed individuals are  equivocal with respect to any causal association  between


 formaldehyde exposure and nasal cancer incidence (18-21). This finding has stimulated


 a  series of research projects aimed at  understanding  the mechanisms involved  in


 formaldehyde-induced toxicity and carcinogenesis.





      ' The genotoxic effects of formaldehyde have been  extensively reviewed (22).


 Formaldehyde induces gene mutations in many organisms including bacteria, fungi,


 yeast, fruit flies, and in cultured  mammalian cells.  Formaldehyde has also been shown


 to induce single-strand DNA breaks, sister  chromatid exchanges, and  chromosomal


 aberrations in a variety of cultured mammalian cells, including human bronchial cells


 (23).
       Formaldehyde induces a variety of toxic effects in experimental animals. It is a
     =*

potent upper respiratory irritant and cytotoxicant that is almost entirely deposited in


the anterior nasal cavity of rodents  (24).   Metabolized  in the nasal  mucosa,


formaldehyde reacts covalently with DNA, RNA and proteins.  The covalent reactions of


formaldehyde with macromolecules are generally accepted as the fundamental causes of


its toxic effects (24).

-------
      In, the  chronic  formaldehyde bioassay  (17),  the relationship between the
incidence of nasal tumors in rats and the concentration of formaldehyde was distinctly
nonlinear.  At 2 ppm, no nasal tumors were present, while between 5.6 and 14.3 ppm,
the tumor incidence increased 50-fold as exposure concentration rose less than 3-fold.
Other concentration-dependent responses in rats exposed  to formaldehyde include
inhibition  of mucociliary function, cytotoxicity, inflammation, and  induction of DNA-
protein cross-links (24-25).

        Molecular dosimetry studies in rats  exposed  to  a range of formaldehyde
concentrations using a DNA-protein cross-linking assay, have shown that formaldehyde
induces  cross-links in rat nasal  respiratory mucosa  following exposure to > 2 ppm
 formaldehyde (26). The rate of formation of these cross-links is a nonlinear function of
 the airborne formaldehyde concentration, increasing more rapidly at high than at low
 concentrations.  The yield of cross-links at a given exposure concentration is probably
 determined by the ability of host defense mechanisms, such as metabolism and DMA
 repair, to maintain the integrity of the DNA. Companion studies by Casanova and Heck
 d'A (27), investigating glutathione mediated metabolism of formaldehyde, have shown
 that saturable metabolism of formaldehyde is an important defense mechanism against
 the  formation  of cross-links.  Currently,  DNA-protein  cross-links  are used as a
 measure of formaldehyde dose at the site of tumor formation (28).
  Formaldehyde-Induced Cell Proliferation

         Increased nasal epithelial cell  proliferation, another  important  biological
  response of the rat exposed to formaldehyde, is a sensitive indicator of respiratory
  epithelial cell toxicity (29).  Increased cell proliferation in response to formaldehyde
  exposure has also been demonstrated in nonhuman primates (30) and xenotransplanted

-------
human  nasal  respiratory epithelium (31).   Little is  known,  however, about the



mechanisms of these  proliferative  responses,  which  may  involve  autocrine and



paracrine growth factors, mutation in growth regulatory genes,  and/or regenerative



stimuli brought about by death of adjacent cells (6, 32-34)







        Epithelial injury with consequent hyperplasia is a common  feature of many



chemically-induced toxic  responses,  including those induced by formaldehyde gas.



Following cytotoxic insult, epithelium of the upper respiratory  tract, including the



nasal passages,  may undergo a dynamic series of alterations to include hyperplasia,



metaplasia,  dysplasia, carcinoma  in  situ or intraepithelial neoplasia, and carcinoma



(35).  These  alterations  are all characterized by a common feature, increased cell



proliferation (35).







       Recent studies have  demonstrated that formaldehyde-induced lesions and



increases in cell proliferation in rats  following acute (1, 4, 9, days or 6 weeks), or



subchronic (3  month) exposure were concentration-dependent.  Nasal epithelial lesions



occurred in specific  regions of the anterior nasal  passages, primarily the walls of the



lateral meatus, mid-septum, and medial aspect of the maxilloturbinate (29, 36). The



increases in nasal  cell proliferation were  associated with formaldehyde-induced nasal



lesions which  included epithelial degeneration, necrosis, hyperplasia, and squamous



metaplasia.
       Increased nasal cell proliferation was present in rats exposed to 6,  10, or 15



ppm formaldehyde following up to 6 weeks of exposure, while no increases were detected



in  the 0.7 or  2  ppm groups  (29).   After 3  months  exposure, increases in cell



proliferation were confined to only the 10 and 15 ppm groups (14).  These short-term



studies demonstrated that 0.7 and 2 ppm formaldehyde do not induce increases in nasal

-------
cell proliferation and that 6  ppm  formaldehyde induces  transient increases in cell
proliferation that return to control  levels by  3  months. The transient increase in cell
proliferation observed at 6 ppm emphasizes the importance of evaluating multiple time
points during a cell proliferation study.  Time-course data on site-specific increases in
cell  proliferation provide important information which is necessary for certain
biologically-based risk assessment models of formaldehyde carcinogenesis (37).

       Various metaplastic-dysplastic or preneoplastic lesions occur in the respiratory
 epithelium of both laboratory animals and humans following exposure to carcinogens
 (38-39). Preneoplasia is generally  believed to  be a precursor response that has a high
 probability of developing into  neoplasia.   Formaldehyde exposure  induces putative
 preneoplastic lesions  in rat nasal  epithelium  following exposure to  carcinogenic
 concentrations (15 ppm) for several months (40-41).  The preneoplastic lesions were
 characterized  by epithelial hyperplasia-metaplasia with  atypia,  similar to those
 reported in extranasal respiratory epithelium (38).

        Cell  proliferation rates in formaldehyde-induced preneoplastic lesions were
 significantly  higher than those of control nasal epithelium  (40).  Labeled cells were
 present throughout the lesion, including the superficial layers, suggesting either a rapid
 migration from the basal cell population to the surface,  or the capability of cells
 throughout  the  lesion to undergo replicative  DMA  synthesis  (40).   Data  on
 formaldehyde-induced  putative  preneoplastic lesions, such  as number of cells per
  lesion,  cell proliferation rate  per lesion, and other parameters, can be utilized  in
  multistage  models of carcinogenesis,  such  as the Moolgavkar-Venzon-Knudson model
  (37).
                                                                                  8

-------
       The tumor incidence from the long-term  formaldehyde cell proliferation study
 (14) confirmed the previously reported bioassay tumor response (17). Similar to the
 bioassay (17), the majority of tumors were squarnous cell carcinomas and arose from
 the nasal epithelium lining the walls of the anterior lateral meatus and the nasal septum
 (14), the same locations reported in other chronic formaldehyde studies (42-43) and
 the sites evaluated for alterations  in cell proliferation.   Nonneoplastic nasal  lesions
 following   long-term  exposure were  concentration-dependent,  and  included
 inflammation, epithelial  hyperplasia, squarnous  metaplasia,  and  necrosis  with
 exfoliation.

       There were no detected treatment-induced responses in cell proliferation in the
 three lowest formaldehyde concentration groups, 0.7, 2, and 6  ppm, following  exposure
 for up to 18 months.  Increases in cell proliferation were present only in the 10 and 15
 ppm groups, and were generally greater in  the 1 5 ppm group as compared  to 10 ppm
 (14).   At each time point evaluated (3,  6, 12,  and 18 months), a good  correlation
 existed between the concentration-dependent increases  in cell proliferation and the
 tumor incidence,  supporting the proposal that sustained  increases in cell proliferation
 are an important component of formaldehyde carcinogenesis (14).
       Site-specific  increases in cell proliferation  following long-term exposure to
formaldehyde were not only present at the lateral meatus and nasal septal locations, but
also the medial maxilloturbinate  (MMT) site,  even though the number of tumors
originating from this site was disproportionately lower as compared to the other sites at
risk. This discrepancy may be attributed to a decreased probability of cell mutation and
subsequent cancer, due to the significantly smaller MMT target area and cell population
at risk.  Site specific nasal  responses could also be  due to differences in regional
susceptibility to formaldehyde,  or other, as yet unidentified, factors.

-------
        The association of epithelial cytotoxicity, cell proliferation and nasal cancer has
 also been demonstrated in a study where male rats with damaged or undamaged nasal
 mucosa were exposed to 10 PPm formaldehyde (44). The nasal damage was induced  by
 bilateral intranasal electrocoagulation of the anterior third of the  nasal passages.  Rats
 with damaged nasal mucosa exhibited increases in formaldehyde-induced rhinitis,
 hyperplasia,  and metaplasia of the nasal epithelium. Exposure to  10 ppm formaldehyde
 for 28 months produced an 8-fold increase  in nasal squamous cell carcinomas in rats
 with damaged noses then in those with intact noses (i.e, not  pretreated with  nasal
 electrocautery but  similarly exposed).  These researchers concluded that both severe
 damage to the nasal mucosa and hyperoroliferation are important in the development of
 nasal tumors in rats exposed to formaldehyde.

  Cell Proliferation and Formaldehyde Risk Assessment

         Cell death and renewal  are predominant features of most toxicologic injuries to
 .the respiratory epithelium.  Toxicant-induced cell necrosis, followed by regeneration,
  could, therefore,  be  a  major determinant  in chemically-induced respiratory tract
  carcinogenesis.  The studies of cell proliferation and nasal epithelium in formaldehyde-
  exposed rats demonstrate a good correlation of cellular injury and cell  proliferation.
  The proliferate and tumor response are dependent  on formaldehyde concentration.
  Induction of nasal carcinoma  in rats by formaldehyde requires long-term.exposure to
'  high concentrations that result in  cell death, followed by regenerative hyperplasiajand
   metaplasia,   changes associated with increases in  cell proliferation.  Since  cell
   proliferation is clearly  involved  in  chemical carcinogenesis, these concentration-
   responsive changes represent potentially important data that could be included in the
   risk assessment process.

                                                                                10

-------
        A pharmacokinetic model, utilizing  the rate of formaldehyde-induced DNA
 protein cross-link  formation,  has been described  in which  the concentration  of
 formaldehyde-induced cross-links formed in corresponding tissues of different species
 can be predicted by scaling certain parameters (45).  DNA protein-cross links are not
 the only  biological  factor affected  by formaldehyde exposure,  however,  and a
 biologically-based risk assessment strategy for inhaled  formaldehdye has been proposed
 (37).  The biologically-based model incorporates molecular dosimetry and cell kinetic
 data, since both of these factors are causally involved in formaldehyde-induced rodent
 nasal carcinogenesis (37).

       Over the  past decade, mechanistic studies  have  provided a great deal of
 information on the pathogenesis of formaldehyde-induced nasal carcinogenesis.  This data
 is important for the assessment of human risks at formaldehyde concentrations below
those  at which cancer develops in rodent bioassays.  Quantitative risk assessment
methods should improve our. understanding of the shape of the formaldehyde nasal cancer
exposure-response  curve, and  the quantitative importance of cell proliferation and
mutation in this process.  Moreover, the incorporation of mechanistically based data will
improve low-dose, and interspecies extrapolation of formaldehyde risk.
                                                                             11

-------
References

1.   Pitot HC. Fundamentals of Oncology, 3rd ed. New York, Marcel Dekker Publ., 1986.

2.   Levine AJ, Momand J,  Finlay CA. The  P53  tumor suppresser gene. Nature
     351:453-456,1991.

3.   Grisham  JW,  Kaufmann WK, 'Kaufman DG.   The cell  cycle  and  chemical
     carcinogenesis.  Surv Syn Pathol Res 1:49-66, 1983.

 4.   Farber E, Sarma DRS.  Hepatocarcinogenesis: A dynamic cellular perspective. Lab
     Invest 56:4-22, 1987.

 5.   Russo J, Russo IH. Biological and molecular bases of mammary carcinogenesis.  Lab
     Invest 112-137, 1987.

 6.   Loeb LA. Endogenous carcinogenesis: Molecular  oncology into the twenth-first
      century. Cancer Res 49:5489-5496, 1989.

 7.   Ames BN,  Gold LS. Too many rodent  carcinogens:  mitogenesis  increases
      mutagenesis. Science 249, 970-971, 1990.

  8.   Butterworth BE, Goldsworthy TL. The role  of  cell proliferation  in multistage
      carcinogenesis.  Proc Soc Exper Biol Med 198:683-687, 1991.

  9.   Cohen SM, Ellwein LB. Genetic errors,  cell  proliferation, and carcinogenesis.
      Cancer Res 51:6493-6505, 1991.

  10.  Preston-Martin S,  Pike MC, Ross RK, Henderson  BE. Cell division and human cancer.
       Prog Clin Biol Res 369:21-34, 1991.

  11.  Chemicals and industrial  processes  associated with cancer in  humans.  IARC
       Monographs, Suppl 1. Lyon,  France: International Agency for Research on Cancer,
       1979.
                                                                             12

-------
 1 2.  Butterworth, BE. Consideration of both genotoxic and nongenotoxic mechanisms in
     predicting carcinogenic potential.  Mut Res 239:117-1 32, 1990.

 13.  Eldridge  SR,  Tilbury LF,  Goldsworthy TL,  Butterworth BE. Measurement  of
     chemically induced cell proliferation in rodent liver and kidney: a comparison of
     5-bromo-2'-deoxyuridine and [3-H]thymidine administered  by injection  or
     osmotic pump. Carcinogenesis 11:2245-2251, 1990.

 14.  Monticello TM. Formaldehyde-induced pathology and  cell  proliferation.  PhD
                                                    9-
     Dissertation, Duke University,  Durham, NC, 1990.

 15.  Swenberg JA, Kerns VVD,.Mitchell Rl, Gralla EJ, Pavkov KL: Induction of squamous
     cell carcinomas of the  rat nasal cavity by  inhalation exposure to formaldehyde
     vapor. Cancer Res 40:3398-3402, 1980.

 1 6.  Albert RE, Sellakumar AR, Lashkin S, et. al. Gaseous formaldehyde and hydrogen
     chloride induction of nasal cancer in the rat.  J Natl Cancer Inst  68:597-603,
     1982.

 1 7.  Kerns WD, Pavkov KL,  Donofrio DJ, Gralla  EJ, Swenberg JA.  Carcinogenicity of
     formaldehyde in rats and mice after long-term inhalation exposure. Cancer Res
     43:4382-4392, 1983.

 1 8.  EPA. Formaldehyde risk assessment update, final Draft. Office of Toxic Substances,
     US Environmental Protection Agency, Washington, DC, 1991.

 19.  Universities Associated for Research and Education in Pathology, Inc. Epidemiology
     of chronic occupational exposure to formaldehyde: Report of the ad  hoc panel on
     health aspects of formaldehyde. Toxicol. Ind. Health 4, 77-90, 1988.

20.  Starr TB,  Gibson JE,  Barrow CS  et al. Estimating  human  cancer risk from
     formaldehyde: Critical  issues. IN "\f. Turoski (ed)., Formaldehyde:  Analytical
     Chemistry and Toxicology, Advances in  Chemistry  Series No. 210, pp 300-333.
     Washington DC: American Chemical Society, 1985.
                                                                           13

-------
21. ' Blair A, Stewart M, O'berg W, et al. Mortality among industrial workers exposed to
     formaldehyde. J Natl Cancer Inst 76:1071-1084, 1986.

22.  Ma TH, Harris MM. Review of the  genotoxicity of formaldehyde. Mutation Res
     196:37-59, 1988.

23.  Grafstrom RC, Fornace A, Autrup H, Lechner JF, Harris CC. Formaldehyde damage
     to DMA  and inhibition  of  DMA   repair in  human bronchial cells. Science
     220:216:218,1983.

24.  Heck H'dA, Casanova M, Starr TB.  Formaldehyde toxicity - New understanding.
     Critical Rev Toxicol 20:397-426, 1990.

 25. Swenberg JA, Barrow CS, Boreiko  CJ, et.  al. Nonlinear biological responses to
     formaldehyde  and  their  implications  for  carcinogenic, risk  assessment.
     Carcinogenesis 4:945-952, 1983.

 26.  Casanova M,  Deyo DF, Heck Hd'A. Covalent binding of inhaled formaldehyde to DNA
      in nasal  mucosa  of F344  rats:analysis of formaldehyde and  DNA by  high-
      performance  liquid  chromatography  and  provisional   pharmacokinetic
      interpretation.  Fundam Appl  Pharmacol 12:397-417, 1989.

'27.  Casanova M,  Heck Hd'A. Further studies on the  metabolic  incorporation and
      covalent binding of inhaled [3H]- and [14C]formaldehyde in F344 rats:effects of
      glutathione depletion. Toxicol Appl Pharmacol 89:105-1 21,1987.

  28.  Environmental Protection Agency. Formaldehyde Risk Assessment Update, June 11,
      1991. Washington, DC:  Office of Toxic Substances, US EPA, 1991.

  29. Monticello TM, Miller FJ, Morgan  KT. Regional increases  in rat nasal epithelial
      cell proliferation following  acute and  subchronic inhalation of formaldehyde.
      Toxicol Appl Pharmacol  111:409-421, 1991.

  30.  Monticello TM, Morgan KT, Everitt Jl, Popp JA. Effects of formaldehyde gas on the
       respiratory tract of rhesus monkeys. Am J Pathol 134:515-527, 1989.
                                                                             14

-------
 31.  Klein  Szanto  AJP,  Ura  H,  Resau  J.  Formaldehyde-induced  lesion  of
      xenotransplanted nasal respiratory epithelium. Toxicol Pathol 17:33-37 (1989).

 32.  Pardee AB.  Biochemical and molecular events regulating cell proliferation. J
      Pathol 149: 1-2, 198(5.

 33.  Lutz WK, Maier P. Genotoxic and epigenetic chemical carcinogenesis: one process,
      different mechanisms. Trends in Pharmacol Sci 9:322-326, 1988.

 34.  Recio  L, Sisk  S, Pluta L,  et. al.  p53  mutations in formaldehyde-induced nasal
      squamous cell  carcinomas in rats.  Cancer Res 52:611 3-611 6, 1992.

 35.  Klein-Szanto AJP. The role  of chemically induced epithelial hyperplasia in the
      development of human cancer. Prog Clin Biol Res 369:35-41, 1991.

 36.  Morgan  KT, Kimbell  JS,  Monticello TM,  Patra AL,  Fleishman A.  Studies  of
      inspiratory airflow patters in the nasal  passages of the F344 rat and  rhesus
      monkey  using nasal molds: Relevance to formaldehyde toxicity. Toxicol Appl
      Pharmacol 110:223-240, 1991.

 37.   Conolly RB, Monticello TM, Morgan KT,   Clewell HJ,  Anderson ME.  A biologically-
      based  risk assessment strategy for inhaled formaldehyde. Comments Toxicol 4:269-
      293, 1992.

 38.   Nettesheim P,  Klein-Szanto AJP,  Marchok AC et. al.   Studies  of neoplastic
      development in  respiratory  tract epithelium.  Arch Pathol Lab Med 105:1-10,
      1981.

39,  Klein-Szanto AJP, Topping  DC,  Heckman  CA, Nettesheim -P. Ultrastructural
     characteristics of carcinogen-induced dysplastic changes in tracheal epithelium.
     Am J Pathol 98:83-100, 1980.

40. Monticello TM, Morgan KT. Cell kinetics and characterization of "preneoplastic"
     lesions in nasal respiratory epithelium of rats exposed  to formaldehyde.  Proc
     Amer Assoc Cancer Res, 30:195, 1989.
                                                                           15

-------
41. ' Morgan KT, Monticello TM.  Formaldehyde toxicity: Respiratory epithelial injury
     and repair. IN Biology, Toxicology and Carcinogenesis of Respiratory Epithelium,
     DG Thomassen and P Nettesheim, eds. pp. 155-171, Hemisphere Publ, New York,
     1990.

42  Morgan KT, Jiang XZ, Starr TB, Kerns WD.  More precise localization of nasal tumors
     associated with  chronic exposure of F344 rats to formaldehyde gas. Toxicol Appl
     Pharmacol 82:264-271, 1986.

43  Woutersen RA, Feron VJ. Localization of nasal tumors in  rats exposed to acetaldehyde
     or formaldehyde. IN: Nasal carcinogenesis in rodents: relevance to human health risk
     (VJ Feron and MC Bosland, eds.), Pudoc, Wageningen, 1989.

 44  Woutersen  RA,  van Gardeen-Hoetmet A, Bruijnjes JP,  Zwart A,  Feron VJ. Nasal
     tumors in rats after severe injury to the nasal mucosa and prolonged exposure to  10
      ppm formaldehyde. J Appl Toxicol 9:39-46, 1989.

 45   Casanova M, Morgan KT,  Steinhagen  WH et.  al.  Covalent binding of  inhaled
      formaldehyde to DNA in the  respiratory tract of rhesus monkeys: Pharmacok.net.es,
      rat-to-monkey  interspecies scaling, and extrapolation to man. Fund Appl Toxicol
      17:409-428, 1991.
                                                                              16

-------
                                    BIOGRAPHY

        James E. Trosko, Ph.D.,
        Department of Pediatrics/Human Development
        College of Human Medicine
        Institute of Environmental Toxicology
        Michigan State University
        East Lansing, Michigan.


        Having received his Ph.D. from Michigan State University in 1963 in
        the area of radiation genetics, Dr. Trosko spent three years as a
        postdoctoral fellow at Oak Ridge National Laboratory, working under
        Drs-. Sheldon Wolff, Ernest Chu and Richard Setlow in the areas of
        radiation-induced  DNA damage/repair  and -mammalian  mutagenesis
        After initiating an academic career at Michigan State University in
        1966, he became  a NCI-Career Development Awardee (1972-1977)  . This
        carcinogenesis  at   the  McArdle Lab  for Cancer  Research at  the
        University of Wisconsin under Dr. Van R. Potter.  He rose throuqh
        the academic ranks at MSU to  Full  Professor,  and along the way was
        awarded the Teacher-Scholar Award, the Distinguished Faculty Award
        Sigma  Xi Senior  Scientist  Award,  and  Outstanding Educator  of
        America Award.   Most recently he was given the opportunity to serve
        as Chief of Research at the Radiation Effects Research Foundation
        in Hiroshima, Japan (1990-1992).   He  is  the author or co-author of
        over 200 journal and book articles.-
_

-------
        PHE ROLE OF MODULATED GAP JUNCTIONAL INTERCELLULAR
             COMMUNICATION IN EPIGENETIC TOXICOLOGY

   Abbreviated title:  GAP JUNCTIONS  IN EPIGENETIC TOXICOLOGY

                    JAMES  E. TROSKO, PH.D.1'3
                     CHIA-CHENG CHANG,  PH.D1
                    BURRA V. MADHUKAR,  PH.D.2
Department of Pediatrics  and Human Development
College of Human Development
Institute of  Environmental Toxicology
Michigan State University
East Lansing, Michigan 48824

'Department of Pharmacology and Toxicology
Indiana Medical  School
Indianapolis, Indiana  46202

Corresponding author:
Department of Pediatrics  and Human Development
College of Human Development
 Institute of Environmental Toxicology
 Michigan State University
 East Lansing, Michigan 48824

-------
ABSTRACT




     The  normal development and health of all multi-cellular



organisms, including the human being, depend  on  the adaptive



maintenance of the integrity of the genetic information  (e.g.,



DNA protective and repair mechanisms), as well as of the homeo-



static and cybernetic regulatory systems within and between



tissues.  The primary focus of the past and current toxicological



studies and risk assessment practices has been to ascertain and



predict the "genotoxicity" of various physical and chemical agents.



The paradigm of "carcinogen as mutagen," while valuable for



stimulating studies of the detection of mutagens and of their



potential role in "causing" somatic and germ line diseases, has



tended to blunt research on the role of non-genotoxic mechanisms



in disease causation.




     This brief analysis will emphasize the need to consider the



role of modulated gap junctional intercellular communication (GJIC)



in any biological risk assessment mode.  It is based on the



following assumptions and facts.   Since gap junctions exist in all



metazoans", they have been associated with the regulation of cell



proliferation, development, differentiation and the adaptive



function of both excitable and non-excitable coupled cells.  A



family of highly evolutionarily conserved genes codes for proteins



(connexins), which, as hexameric units (connexons),  form membrane



associated channels.  Cells coupled by gap junctions will have



their ions and small regulatory molecules equilibrated.  Regulation



of GJIC can be at the transcriptional,  translational or post-



translational levels.   Transient down or up regulation of GJIC can




                               2

-------
be induced by endogenous or exogenous chemicals via many mechanisms
at any of these three levels.  Stable abnormal regulation has been
associated with activated oncogenes,  and normal regulation has been
associated with several tumor  suppressor genes.
     The  dysfunction of  these gap junctions might play a role in
the   actions  of  various   toxic   chemicals   which  have  cell
type/tissue/organ  specificity.  This  could bring about distinct
clinical  consequences, such  as embryo lethality or  teratogenesis,
reproductive dysfunction  in  the gonads,  neurotoxicity of the CNS,
hyperplasia  of the skin,  and tumor promotion of initiated  tissue.
Modulation of  GJIC   should  be viewed  as  a  scientific basis of
 •epigenetic  toxicology'  since  the  alteration of  intercellular
 communication  would  alter the internal physiological state of  the
 cell.    The  inhibition  of  GJIC  is  a  necessary  component of
 mitogenesis  (a necessary component of the multi-stage carcinogenic
 process).  The modulation of GJIC can have both toxicological, as
 well as  therapeutic potential.
 KEY WORDS:     Gap junctions; epigenetic tocicology; intercellular
                communication;  connexins; oncogenes

-------
 •During evolution, long-lived multi-cellular organisms must have
 developed  defense  mechanisms   to  protect   them   against  the
 carcinogenic  and  other   deleterious  effects   of   spontaneous
 mutations	Protagonists  of the theory of cell replication leads
 to  cancer do not deal with, this aspect or explain how this barrier
 might  be broken during  tumor development.'
                                    I.E.  Weinstein (1)
 INTRODUCTION;   BIOLOGICAL  BASTS  POR RISK ASSESSMENT:  WHAT FACTORS
                ARE.... INVOLVED?
     Implicit  in the acute and chronic  exposure to radiations  and
 chemicals is a risk to normal short and long-term functioning of a
 multi-cellular  organism and  to its offspring.  Predictions  of  the
 potential "toxicities- are being made on the bases of  mathematical
 models based on incomplete empirical or epidemiological data, known
 or suspected mechanisms of  action of the agent on a single level of
 biological  organization and extrapolations from a  variety  of
 laboratory animal studies (2) . In  lieu of complete understanding of
how these toxic agents  might act  in  the human being  (which will
never be possible), we are  left with the challenge to develop risk
assessment  models  that, at   least,  begin  to approach  having a
biological basis (3).
     The quotation  of I.E. Weinstein serves to illustrate the point
that,  as  toxicologists,  we have a long  way to go.  On one end of
the problem,  we must recognize that a multi-cellular organism, such
as the  human being, is  not just  a  collection of  10" independent
cells in  a hairy-covered bag made of  skin.   We  are the, result  of
the hierarchical process  (4) of cybernetically interacting elements
                                4

-------
(5) .  Fundamentally, the health and adaptive ability (homeostasis)
of a multi-cellular organism is based on a number of systems on the
molecular,   biochemical,    cellular,   tissue,    organ,    system
(physiological, immune, nervous)  levels (4). Biological organisms,
both single and multi-cellular, have developed a series of adaptive
mechanisms  at  each  of  these  levels  to  survive  the  constant
exposure, either acute or chronic, of radiations and chemicals. The
toxic endpoint at the cellular  level  could be  mutation  of the
genome,  cell  death or  epigenetic alteration of  the phenotype  (5)
 [Figure 1].
                GENOTOXICANT
                                EPIGENETIC TOXICANT
                          CHROMOSOMAL
                            MUTATION
                                                   ©
                                               TRANSCRIPTIONAL
                                                MODIFICATION
                                               TRANSLATIONAL
                                                MODIFICATION
                                              POSTTRANSLATIONAL
                                                MODIFICATION
       Fig.  1    Diagram  illustrating  the difference between
       gXotoxic and epigenetic chemicals.  Those ^at alter the
       Duality or quantity of genetic information are
       while those  that affect  the  expression  of the
       information,  at the  transcriptlonal,  W"81*
       posttranslational   levels   are   epagrenetic
        (Reprinted from Trosko et al, In:  In Vitro
       Toxicology,  G.  Jolles  and  A.  Cordier,  eds,  Academic
       Press, NY, 1992; used with permission)

-------
     The  presence  of  melanin   in   the  skin .  tissue  or  drug-
metabolizing  enzymes  in  the  cell  could  protect  the  DNA  fronr-
ultraviolet light   or  chemical induced  damage,  respectively. DNA
repair enzymes have the potential  to  either restore the genome to
its original condition,  or, if not repaired or repaired in an error-
prone  fashion,   can lead  to  a  viable  or  non-viable  mutation.
Depending of the nature of the  mutation,  the gene mutated, whether
the  mutated  gene  is  expressed,  whether  the  mutated  cell  is
replicated, the mutation might  have little  biological significance
because of cellular redundancy  or it might have devastating somatic
or hereditary consequences (6)  .  At the  tissue level, cell death,
depending on the number of cells, which cells die, at what stage of
development, might or might not affect the  homeostatic or adaptive
status of the multi-cellular organism.   For example,  the death of
one heptacyte might not even  elicit a  regenerative' response of the
liver.   On  the  other  hand,  the death of  the single stem or
progenitor cell of a critical organ during  development could have
lethal or devastating teratogenic effects.  The death of many cells
could  induce  compensatory  or  regenerative  hyperplasia,  which,
depending on  the cause of  the death  [apoptosis  (7,8)  or induced
cytotoxicity], lead to natural tissue restructuring, wound healing,
scar tissue formation  or  tumor promotion  (2).   At the  organ and
system levels, control  of both cell growth  and differentiation is
mediated by extracellular communication mechanisms [immune systems,
neuro-endocrine  systems]  (9) .  In  other  words,  even if a  viable
mutation occurs,  in a  single cell, if  tissue,   organ and  system
suppression mechanisms  prevent  the product of that abnormal cell to

-------
disrupt  homeostasis or  prevent  the  mutated  cell  to  clonally
increase, then these systems act as another protective barrier to
maintain health and homeostasis.
     While the details of each of these barriers at the molecular,
biochemical, cellular and physiological levels have been recognized
and  studied to various  degrees/one  seems to  have  been  largely
ignored in  the field of toxicology  and  risk  assessment. How the
multi-cellular organism  suppresses  the  potential   toxicological
consequences  of  abnormal   cell  proliferation/differentiation of
either normal  or mutated cells will be the topic of  this  analysis.
Specifically,  the  role of intercellular  communication will be
examined  in  the  context  of  acting  as   a  barrier   to  cell
proliferation and how the modulation  of intercellular communication
 could contribute to the disruption of homeostasis. This new concept
 will   form  the  basis  of  the  emerging  field  of  -epigenetic
 toxicology' (10,11).
 Current Problems in Risk Assessment
      Fundamentally, extrapolations from data  derived from in vitro
 and in vivo tests,  as well as from human epidemiological data, are
 dependent  on  both  the theories of the disease endpoints  in which
 one  is interested, the limitations of each  test system (assuming
 for  the moment the  validity of the  design  and  execution of the
 experiment),  and  the assumptions of  the extrapolations  from the
 non-human  test data to the human population/individual.  Limitations
 of  in vitro tests, especially  those  designed  to detect genotoxic
  chemicals   (12-14),  and of  in vivo  bioassays  (15-22)   have  been
  noted.  Epidemiological  approaches  are  characterized  by  poor
                                  7

-------
 sensitivity and the inability to determine  underlying mechanisms.
 Compounding the problem are the facts of  interactions of physical
 and  chemical  exogenous  agents  with  endogenous  factors  having
 genetic, developmental and sex elements  (23).
      Recently  the new  concept  of  'molecular  epidemiology  has
 emerged  (24-26) .   It  is  based on the assumption that  if  a given
 causative  agent   leaves  a unique  molecular  fingerprint   in  the
 diseased tissues/cells/macromolecules/DNA, then one might  predict
 and prevent these diseases (26) .   While there exists some  promise
 to this new approach,  some of the same  problems will plague this
 approach since it  is  based on some  of  the same assumptions  and
 limitations of the previous approaches.
      In order  to limit this discussion, the endpoint of cancer will
 be used to  illustrate the objective of this analysis. However,  it
 should be obvious  from this analysis that other disease" endpoints,
 such    as   teratogenesis,   reproductive   dysfunction,    immune
 dysfunction,   neurotoxicity,  cardiovascular  diseases  and  other
 disease states, will be involved.
 Mutagenesis and Mitooenesis in Carcinoqenesis
      It is  in the  underlying assumption  of  the  mechanism of
 carcinogenesis^ from which much of the problems of risk assessment
 to cancer after exposure to  radiation or chemicals come.  It  is now
 abundantly clear that no one thing "causes" cancer ,(27) .  Pathology
 information and epidemiological data  show that human  cancer  is the
 result  of multiple steps during the evolution of a normal  cell to
a metastatic cell  (28-30) .  While the role to mutations  in  cancer
was  well  known   to  geneticists  via   the  various hereditary
                                8

-------
predispositions   to   cancer    (xeroderma   pigmentosum,   Downs,
retinoblastoma, Wilms,  Fanconi's, etc.), it took some time before
a more general acceptance of the somatic mutation theory of cancer
became  commonly  accepted.  The  introduction  of  the  concept,
-carcinogens as mutagens-  (31), helped to spur the development of
new  interests,  assays,  and  experiments into  the  detection of
"carcinogens'.  Identification  of DNA  lesions, such as pyrimidine
dimers, being correlated with higher mutations and cancers in cells
of   sun-light   induced  skin   cancer-prone   syndromes,  xeroderma
pigmentosum   (32-35),    together  with  the  identification  of
electrophilic-induced   DNA  lesions   in   tissues   capable  of
metabolizing  chemicals   (36),  helped  to  shape  the  idea  that
mutations were responsible for cancer.
      With  the relatively  recent  identification  of  -oncogenes-
 (37,38),  and more  recently the -tumor suppressor-  genes (39,40),
 more evidence was supplied to bolster the somatic mutation theory
 of cancer. With modern molecular technology, it became clear that
 these important oncogenes and  tumor  suppressor genes in tumor cells
 were often mutated. However,  as with  most exciting new scientific
 theories,  the. beginning  euphoria starts to wear thin and the new
 paradigm must address challenges.  These challenges included the
 animal experiments showing the multi-step nature of.carcinogenesis
  (41), being conceptualized by the  -initiation-, -promotion", and
  -progression-  stages.   In   addition,  many  of  the   so-called
  •carcinogens- in the animal  bioassay test were shown to be  non-
  mutagens in various in  vitro  assays presumably designed to detect
  mutagens (42).  Furthermore,  most of  these animal  tumor promoters

-------
 were shown also to be non-mutagenic  in these  in vitro genotoxicity
 assays (12,13) .
      Recently,   the idea  emerged that  since  initiation of  the
 carcinogenic process started  in  a single cell  [evidenced by  the
 irreversibility of the event  in a single  cell  and the monoclonal
 nature of most tumors (43,44)], the ultimate appearance of a multi-
 celled tumor from  this  single initiated cell most have  been  the
 result of  the promotion process  (45) .  In other  words,  promotion
 must be, at least, a mitogenic  process for  the initiated cell  (11) .
 Since  many  tumor promoters  and  promoting conditions (i.e., surgery,
 cell killing  (46-48)] seemed to be associated with hyperplasia  and
 cell proliferation,  it seemed plausible to assume that mitogenesis
 is  a  necessary  component  of  carcinogenesis   (49) .  The idea  was
 further supported by the findings  that many tumor promoters acted
 as non-genotoxic  agents in several in vitro tests for genotoxicity
 (50) and many were known to be mitogenic in either in vitro or in
 vivo systems, such as growth factors  and hormones.  In addition, if
 multiple genetic  'hits' were needed for the carcinogenic process to
 occur, mitogenesis was necessary to complete the process  since, by
 definition,  a  mutation  is  the  hereditary  transmission of  an
alteration  of a change in the genome. Both  spontaneous and induced
mutations  depend on mitogenesis to  'fix"  the alteration in  DNA.
However, as correctly pointed out  by  Weinstein (1), excessive cell
proliferation,  in and of  itself,  is probably  not the causative
 factor in most cancers. The observations that  additional exposures
to  initiating agents are  necessary  to  convert  promoted benign
tumors  to  carcinomas  also   supports  the  notion  that, while
                                10

-------
mitogenesis is  a  necessary component of carcinogenesis  (51-57), it
is an insufficient causative agent (58).
Gao Junctional  Communication in Mitoaenesis and Tumor  Promotion
     In order to  explore  the role of mitogenesis in carcinogenesis,
particularly as to how the tumor promotion concept relates  to it,
the potential role of gap Junctional communication in  mitogenesis
will  be  examined.  The  concept  of  contact  inhibition  (59) was
created  to explain  the  fact  that  most   normal   cells  stop
proliferating, in vitro, when they come in direct contact with each
other.  Cancer  cells  have been  characterized  by their inability  to
contact inhibit  (60,61).  With the demonstration that most, if not
all,  cancer cells exhibit some form of dysfunctional GJIC (62-65),
 it  seemed  logical   to  link  GJIC   with  the  control  of  cell
 proliferation. In addition,  many known endogenous  and exogenous
 tumor  promoting chemicals  have  been shown  to reversibly down
 regulate  GJIC,  either  at the transcriptional,   translational  or
 posttranslational levels (50,66). Even physical  tumor promoting
 conditions, such as  partial hepatectomy (67),  have been  associated
 with  the down regulation  of  GJIC during the regenerative  phase
  (68).  "cell killing, which does  release  extracellular mitogenic
  stimulating chemicals  needed  for regeneration,  could  indirectly
  cause  the down regulation of GJIC and bring about a tumor promoting
  effect. Prostaglandins  and their metabolites  have been associated
  with  the lysis of  dead cells  and  the down regulation  of GJIC
  (69-72). Genotoxic  agents,  depending on the dose or concentration
  (73),  could be either initiators or  -complete  carcinogens'. At non-
  cytotoxic  levels,   these  agents might  primarily   initiate.  At
                                  11

-------
 cytotoxic  levels,  they would  initiate  as well  as  promote  the

 surviving  initiated cell. There are also, non-genotoxic chemicals,

 such  as alcohol or heptocytotoxic viruses, which could kill cells,

 thereby   forcing  compensatory   hyperplasia  of  any   surviving

 spontaneous  or induced initiated cell.   Many alcohols have been
                                                 4
 shown to inhibit  GJIC  at near or  at  cytotoxic  levels  (74).

      While the evidence rigorously supporting  the  role  of  GJIC in

 "contact inhibition" and growth  control has not yet been generated,

 there is strong  inferential support  for this hypothesis.  Few, if

 any,   cells   which  are  gap   junctionally  coupled  have  been

 demonstrated to be dividing. In  addition, strong correlations with

 the  lack  of GJIC have been shown with cells which  are not gap

 junctionally coupled  (see  58).   Also,  as previously noted, tumor

 promoting  chemicals, growth factors, or  physical conditions, such

 as  cell  removal  or   killing,  have  been  associated  with both

 mitogenesis of the  initiated cell and reduction of GJIC  (50). The

 observations that some mitogens or  mitogenic  conditions are not

 promoters  (75,76),  is,  in and of themselves,  not  rigorous proof

 that  GJIC  is  not involved.   Unless  it can  be shown  that  the

mitogenic .stimuli is  both sustained (77) and  is influencing the

 initiated  cell,  mitogenic stimuli which are only  transient  and

affecting  only the  non-initiated  cell will  not be  a promoter.  In

addition,  any mitogenic assay which indicated that a given chemical

did not act as a  mitogen in a given  tissue/ organ does not prove

the chemical might not be a tumor promoter since it might not act

as a mitogen for  normal cells but  only for the few initiated cells

in the tissue. The observation that phenobarbital can be a promoter


                                12

-------
in the liver is not determined by the observation that  there exists
a sustained hyperplasia in  the  liver,  but  the selective clonal
expansion of  the initiated  cells.  The fact  that  "apoptosis'  is
inhibited by  several  agents  that  are tumor  promoters raises  an
interesting concept that  the selective accumulation of initiated
               *
cells may well be the end result of the blockage of  GJIC needed for
controlled  or programmed  cell  death  (78) .   In other words,  in
closed organs, such as  the  liver, apoptosis is a necessary  means to
control  the volume of the  organ. A balance between cell birth and
cell   death might be necessary  to  maintain the stable number of
cells  (79) . The inhibitions of GJIC could affect either or both the
control  of  cell  division or cell death.
      Another  line of evidence linking mitogenesis  with GJIC  comes
 from the field  of  oncogenes and  tumor  suppressor genes. Proto-
 oncogenes and their activated counterparts, oncogenes, are defined
 as  those genes controlling cell  proliferation and  differentiation
 (37,38).   On the other  hand,  tumor  suppressor genes are  those
 normal  genes  which,  by  definition,  prevent  cell  proliferation
 (39,40).  A number of  oncogenes  have now been associated with  the
 down-regulation  of   GJIC   and   cell   proliferation  (80-92) .
 Furthermore, several tumor  suppressor genes have  been associated
 with the up-regulation of GJIC (93,94).
      In  recent  years,   several anti-tumor promoters  or  anti-
 carcinogens have also  been associated with the up-regulation of
 GJIC. Retinoids (95-99),  c-AMP (100-103),  carotenoids  (104)  and
 lovastatin (105) have been linked with increased GJIC  and  decreased
 cell growth and restoration of a normal phenotype.
                                 13

-------
      If in fact the lack of GJIC is causative of a lack of growth

 control and cancer, then it would be predicted that a restoration

 of GJIC in  non-communicating,  tumor cells  should  lead  to growth


 control and a  normal  phenotype.  Transfection  of several  non-

 communicating tumor cells has led to the restoration of GJIC and,

 at least,  partial growth control in vitro and in vivo (106-108) .

 Stem Cells,—Gap	Junctional Communication,  Differentiation  and

 Carcinoqenesis


      One of the  critical assumptions to  be made in cancer  risk

 assessment involves the  problem,  -Which cells are the target cells

 for  the  carcinogenic  process?'   Is  any  cell  in  all  tissues

 potentially capable of being converted to a cancer cell?   Or,  is
                          4>
 there only a few special types  of  cells capable of tumorigenic

 transformation? One theory, namely, 'oncogeny as partially blocked

 ontogeny  (109) , supported by the theory of 'cancer as a  disease of

 differentiation'  (110,111),  leads  one to hypothesize  that  stem

 cells are the  target cells.  Since stem cells are defined as those

 giving  rise  to  one cell that goes down a  differentiation  pathway

 and another  cell  retaining stem  cell properties.


     Evidence showing only certain cells  are transformable came

 from  the observations by T'so and  colleagues  (112);  They  showed

 only a few  'contact insensitive'  cells were the ones which could be

 transformed  by carcinogens.  Since cancer cells are characterized

by,  not only  their  lack of  growth  control,  but  also  by  the

 inability  to terminally  differentiate under  normal  conditions.

Cancer  cells appear to  be  stem cells  having  the  ability  to

proliferate but unable to control their growth or to differentiate.


                               14

-------
The ultimate stem cell,  the fertilized egg or  zygote  appears to
lack GJIC. Only  after  the early  stages of  development,  do these
cells express various gap junction genes  (113),  thereby creating a
cellular mechanism for  growth control,  tissue compartmentalization
and  differentiation.   Using this idea,  normal human  kidney and
breast epithelial stem cells have been isolated from human tissue
 (114,115).   The  question now arises that if normal stem cells do
not  have  GJIC  and cancer cells do not have GJIC,  then why aren't
normal stem  cells cancerous? The  answer appears to be that normal
stem cells can be induced to express GJIC very easily and they then
become  progenitor  and  differentiated  cells.    Moreover,   they
 probably  control their  cell proliferative activity by extracellular
 communication mechanism, that is,  extra-cellular  negative growth
 regulators,  such as  TGF-fr from the differentiated daughter cells,
 might suppress'stem  cell growth  (116-119).  If either the negative
 growth regulator is  reduced or mutated or that the receptor on the
 stem cell is  reduced  or mutated, the stem cell could  grow  in an
 uncontrolled manner. If  the stem cell  cannot regulate its GJIC, it
 probably  cannot  differentiate   properly.  These  might  be  the
 biological  control points affected  by the carcinogenic process.
      A very significant  observation has been recently made, in that
 tumorigenic and non-communicating  glioma  cells  could  be growth-
 arrested when co-cultured with sister  glioma cells transfacted with
 a connexin43 gene (120)  .  These transfected glioma cells, with the
 expressed connexin43, were able  to establish GJIC.  However, the
 growth arrest of the tumorigenic  and non-communicating glioma cells
 was via  a  soluble  factor, produced  by the newly  communicating,
                                  15

-------
 transfected  glioma cells,  not by  GJIC between the  tumorigenic
 glioma cells  (which do not have GJIC)  and the  transfected glioma
 cells.   In  other words, growth arrest  was via  an  extra-cellular
 communication  mechanism   (negative  growth  regulator)  between
 heterologous  cells,   produced- by  inter-cellular  communication
 between homologous cells.
      The implication of this study could explain (a) the selective
 nature of metastatic  cells, and  (b)  the  growth control  of  non-
 communicating stem cells by communicating differentiated daughter
 cells.  In the former  example,  if a non-communicating tumor  cell
 lands in  a  distal tissue  which  produces  an  effective negative
 growth regulator,  the  tumor cell  would not  grow.  On  the other
 hand,  if the metastatic cell invades  a tissue  that lacks a negative
 extracellular growth  regulator, it would continue to pr-oliferate.
 This  could explain  the  "seed and soil- concept of tumor metastasis
 (121).
      In  the  latter case, a normal non-communicating  stem cells,
 surrounded by communicating differentiated daughters could produce
 a  negative extra-cellular growth regulator keeping  the  stem  from
 dividing.   Removal  or  blockage  of   the  source  of  the negative
 extracellular  growth  regulator would  allow  the stem  cells to
proliferate and differentiate.
     If the  preceding  hypothesis  is  correct,  then the number of
stem cells in a given tissue and during the developmental  and aging
process. The  question  which needs to be  answered  is,  -Are  the
number of stem cells in  all tissues the same  during development and
aging?" For  example,  open-ended  tissues,   such  as  the  skin  and
                                16

-------
lining  of  the 'GI  tract,  have  their  stem  cells  constantly
proliferating until death.   On the other hand, organs, such as the
lung and testes, must maintain  their volume as well as replace lost
cells. Other organs, such as the  liver and kidney,  must primarily
maintain  the  volume of the organ once  they  finish  their growth.
The human breast might provide the evidence  linking the stem cell
as  the target cell for carcinogenesis. Evidence for ' the risk to
cancer in the survivors to a bomb radiation seems to be highest in
young women who developed breast cancer  (122).   One explanation
might be that the  number  of  stem cells  of  the  breast tissue is
highest in  these women, particularly if they have never conceived.
 Pregnancy would be expected,  in  the breast  tissue to deplete  the
 stem cell pool by  virtue  of converting these stem cells to milk
 producing  terminally  differentiated   cells.   Again,   if  this
 hypothesis  is true,  then  there  might be a need to consider  the
 number of  stem cells  available  for  transformation in  each  tissue
 during  each  developmental  state of the individual.  This  would
 influence the initiation phase of carcinogenesis. One would predict
 that  as one  ages,  the number of stem cells in some  tissues would
 decrease, yet at the same  time,  as we age, the number of initiated
 stem cells in these tissues  would  increase (spontaneously or by
 induction). The critical issue to be considered  then, would be the
 amount of  promotion that  occurs after initiation.
 SUMMARY; IMPLICATIONS TO  RISK  ASSESSMENT
       Regulation   of  cell  growth,  development/differentiation,
 homeostasis  and  adaptive  responses to physical,   chemical  and
 biological agents in a metazoan can occur at  all  levels of the
                                  17

-------
 biological  hierarchy.  At  the  cell  level,  extra-,  intra-  and

 intercellular  communication mechanisms  help maintain homeostatic

 regulation   of   these  important  functions  [Figure  2].   These
                ENVIRONMENTAL
                 CHEMICALS
               [TPA. DOT.SACCHARIN,
                PHENOSARBCTALI
(HORMONES, SHOWTH FACTORS.
 NEUHOTRANSMITTER3. ETC ]
                             EXTRACELLULAR
                             COMMUNICATION
                                           INTRACELLULAR
                                           COMMUNICATION
                    SAP JUNCTION
                  INTERCELLULAR COMMUNICATION
             © ALTERS MEMBRANE
               FUNCTION

             ® ACTIVATES INACTIVE
               PROTEINS

             © MODULATES SAP
               JUNCTION FUNCTION

             © MODULATES GENE
               EXPRESSION
      Fig.   2    The  heuristic   schemata"'characterized  the
      postulated link between  extracellular communication and
      intercellular  communication  via  various  intracellular
      transmembrane  signalling mechanisms.    It  provides  an
      integrating view of how the  neuroendocrine-immune system
      ("mind or brain/body connection")  and other multi-system
      coordinations  could  occur.    While  not  shown  here,
      activation or altered expression of various oncogenes (an
      "anti-oncogenes") could also contribute to the regulation
      of gap junction function.  (Reprinted from J.E. Trosko and
      C.C. Chang, Toxicology Letters 49:283-295, 1989, Elsevier
      Science Publisher; used with permission)

communication  mechanisms  have   evolved  via   evolution   to  be

intimately integrated,  such that perturbations of one communication

mechanism will affect  the other communication processes (123).  In

other words, hormonal type-extracellular communication can modulate

gap  junctional  intercellular communication  via  alterations  in
                                 18

-------
intracellular  communication  second  messages,   such  as   c-AMP
increases.   The normal homeostatic  control of  these integrated.
communication processes can be disrupted by exogenous factors which
either mimic, in part, the endogenous extracellular communicating
signals or interfere with their ability to act. Stable interference
of this integrative communication process can also occur when the
genetic  information  coding  for  any  of the  three  communication
networks is either mutated,  eliminated by cell removal/cell death
or epigenetically altered.
     The  question  of  thresholds  or lack  of thresholds  needed to
a.lter  the  homeostasis at  any  of  these  levels  (molecular  to
systems),  which could bring about  a  disease state,  needs  to be
answered.   Resolution, of  this problem will  not  be  easy.   On one
hand,  protective systems,  redundant  genetic information  and DNA
repair systems  seem  to suggest  that both  mutagenesis  and cell
killing, as cellular endpoints, would not demonstrate non-threshold
responses.  As  to whether epigenetic  events  could occur with non-
threshold kinetics,  one  can  only speculate.  If the homeostatic
state  of  a cell  in a  multicellular organism is one where the cell
is either in  the G  state with a given set of genes expressed or a
differentiated  cell  which  is   not   responding  to  and adaptive
stimuli,   one can assume that  the  intercellular communication
network  has  not  been  perturbed.  Therefore,   if  intercellular
communication is -modulated, either up or down, one could expect the
cells  to  respond by turning on or off the genes necessary  for cell
proliferation,   cell  differentiation  or adaptive  differentiation
responses.  Evidence  that gene expression is altered by alteration
                                 19

-------
 of. intercellular communication  is  almost self evident.   There  is
 even evidence that  threshold levels of growth factors,  hormones,
 oncogenes and chemical tumor promoters are needed to modulate GJIC
 (92,124-128).
      To complicate  the matter of  the relevance of GJIC to risk
 assessment after exposure to toxic agents is,the role of stem cells
 to  various  disease  states  that  • are  the  result  of   not  one
 dysfunctional cell, but  the consequence of a  dysfunctional cell
 having been amplified  in _a given tissue so as to make its presence
 known to the  whole body by  its ability to" disrupt homeostasis  at
 the  systems level.  The number of  stem cells in  various tissue
 during aging is one crucial parameter and the accumulation of those
 stem cells  which have  been blocked  in their ability to terminally
 differentiate,  but  not in  the  ability  to  proliferate,  will  be
 potential determinants of future disease states.
      The  endpoint  of GJIC should be seriously considered in the
 assessment  of  toxic  agent  exposure.  Abnormal  GJIC  has been
 associated with  a wide variety of disease states.   Since GJIC has
 been associated with development, differentiation and wound healing
 (68,113,129-131), it should  not  be  surprising to note  that  their
 dysfunction has been   associated  with  teratogenesis   (132,133),
 neurotoxicity (134),  reproductive dysfunction (135),  cardiovascular
 diseases    (136,137),   cataract   formation    (138),    ischemia,
hypertension     (139,140),    cholestasis    (130),     hereditary
mucoepithelial dysplasia  (141), as well  as Chagas  disease (142).
These gap junctions exist in  all  tissues of the body. Eac.h type of
gap  junction  protein  is  probably  regulated  differently  and  in
                               20

-------
different cell  types,  the  same  gap junction  might  be regulated
differently due to the physiological states of the different cell
types.  The  interaction of  multiple endogenous  and/or exogenous
chemicals   could,   by  the  net   result   of  that  interaction,
synergistically, additively or antagonistically modulate GJIC. To
ignore  this fundamental structure in the maintenance of homeostasis
of  health  of  the  human  being   will  be  to  ignore  one  of  the
parameters  of a biologically-based risk assessment model.
ACKNOWLEDGEMENTS
      The authors  wish  to acknowledge  the  excellent technical
assistance  of Ms. Heather  Rupp and Mrs. Beth Lockwood  for  some of
the research on which  this paper was based and the skilled  word
processing  of  Mrs.  Jeanne McHugh.   The  research,  on which  this
manuscript  is based, was supported, in part,  by grant from the U.S.
Air  Force  Office  of  Scientific Research  (USAFOSR-89-0325),  the
NIEHS  (1P42ES04911) and the National Cancer Institute (CA21104).
 REFERENCES
   1. Weinstein,  I.E.   Mitogenesis   is   only   one  factor   in
      carcinogenesis. Science, 251: 387-388, 1991.

   2. Trosko, J.E. and Chang,  C.C. Implications for risk assessment
      of genotoxic  and non-genotoxic mechanisms in carcinogenesis.
      in: V.B. Vouk, G.C. Butler,  D.G. Hoel and D.B. Peakall (eds.),
      Methods for  Estimating Risk of Chemical  Injury:  Human and
      Non-Human Biota Ecosystems, pp. 181-200,  Chichester,  England:
      John  Wiley and Sons.  1985.
                                 21

-------
 3. Trosko.   J.E.  and  Chang,  C.C.   The  role  of   inhibited
    intercellular communication  in carcinogenesis: Implications
    for risk  assessment  from exposure  to chemicals.  In:  C.C.
    Travis  (ed.),  Biologically  Based  Methods  for Cancer  Risk
    Assessment,  pp.  165-179, New York: Plenum Press. 1989.

 4. Brody, H.  A  systems view of man:  Implications for medicine,
    science  and  ethics. Perspect Biol Med,  17: 71, 1973.
 5. Potter,  V.R. Probabilistic  aspects  of the  human  cybernetic
    machine. Perspect  Biol Med,  17: 164, 1974.

 6.  Trosko,  J.E. and Chang,  C.C.  The role of  DNA repair capacity
    and  somatic  mutations in carcinogenesis and aging.  In:  H.T.
    Blumenthal (ed.), Handbook of Diseases of Aging, pp. 252-295,
    New  York: Van Nostrand Reinhold Co.  1983.

7.  Wyllie, A.H.  Apoptosis  and  the  regulation  of cell numbers in
    normal and neoplastic  tissues. Cancer Metastasis  Rev,  11:
    95-103, 1992.

8. Williams,  G.T.   Programmed   cell  death:   Apoptosis  and
   •oncogenesis.  Cell,   65: 1097-1098, 1991.
9. Trosko,   .J.E.   and  Chang,   C.C.   Role   of   intercellular
   communication in modifying  the  consequences  of mutations in
   somatic  cells.  In:  D.M.  Shankel,  P.D. Hartman, T. Kada and A.
   Hollaender  (eds.),  Antimutagenesis  and  Anticarcinogenesis
   Mechanisms,  pp.  439-456, New York:  Plenum Publ. 1986.
                             22

-------
10. Trosko,  J.E.,  Chang,  C.C.,  and  Madhukar,  B.V.   In  vitro
    analysis  of   modulators   of  intercellular  communication:
    Implications for biologically based risk assessment models for
    chemical exposure.  Toxicol.  In Vitro,  4:  635-643,  1990.

11. Trosko,  J.E.,  Chang,  C.C.,  Madhukar,  B.V.,  and  Oh,  S.Y.
    Modulators  of gap junction  function:  The scientific basis of
    epigenetic  toxicology. In Vitro  Toxicol.,  3:  9-26,  1990.

12. Trosko,  J.E.  A failed paradigm: Carcinogenesis  is  more than
    mutagenesis.  Mutagenesis,  3: 363-366, 1988.

13. Trosko,  J.E.  A  new  paradigm  is  needed in  toxicological
     evaluation. Environ. Mutagen.,  6:  767-769, 1984.

 14.  Clive,  D.  Genetic  .toxicology:  From  theory  to  practice.  J.
     Clin. Res. Drug Devel.,  1:  11-41,  1987.

 15. Clayson,  D.B. The need for  biological  risk  assessment  in
     reaching decisions about carcinogens. Mutation Research, 185:
     243-269, 1987.                                  .

 16. Douglas,  G.R., Blakey,  D.H.,  and Clayson, D.B. Genotoxicity
     tests as  predictors  of carcinogens:  An  analysis. Mutation
     Research,  196: 83-93,  1988.
                                 23

-------
 17.  Ashby,    j.   Origins   of    current    uncertainties    in
     carcinogen/mutagen  screening.  Environ Mol Mutagen, 16: 51-59,
     1989.

 18.  Ames,  B.N.,  Magaw,  R.,   and  Gold,  L.S.  Ranking  possible
     carcinogenic hazards.  Science,  236:  271-280,  1987.

 19.  Travis,  C.C.,  pichter  Pack,  S.A.,  Saulsbury,   A.W.,   and
     Yambert,  N.W.   Prediction  of  carcinogenic  potency   from
     toxicological data. MUTAT.  RES., 241:  21-36,  1990.

 20.  Goodman, J.I.  Letter  to  Editor.  Molec.  Carcinogenesis,  5:
     247-248, 1992.

 21.  Cohen,   S.M.   and  Ellwein,   L.B.   Cell  proliferation   in
     carcinogenesis. Science, 249:  1007-1011, 1990.

 22.  Ames,  B.N.  and  Gold,  L.S.  Too  many  rodent  carcinogens:
     Mitogenesis increases mutagenesis.  Science,  249:  970-971,
     1990.               ,

23.  Trosko,  J.E. Towards  understanding  carcinogenic  hazards: A
     crisis in paradigms.  J. AM.  COLL.  TOXICOL., 8:  1121-1132,
     1989.
24. Shields, P.G. and Harris,  C.C. Molecular epidemiology and the
    genetics of environmental cancer. J. A. M. A., 266: 681-687,
    1991.    •                             :
                               24

-------
25. Jones,  P.A., Buckley, J.D., Henderson, B.E., Ross, R.K., and
    Pike,  M.C. From gene to carcinogen: A rapidly evolving field
    in molecular epidemiology. CANCER RES.,  51:  3617-3620, 1991.

26. Vogelstein,   B.    and   Kinzler,   K.W.   Carcinogens   leave
    fingerprints. Nature, 355: 209-210, 1992.
27. Trosko, J.E. Does radiation cause cancer? In: RERF Update, pp.
    3-5, 1992.

28. Fialkow,  P.J.  Clonal origin  of human tumors. Am. Rev. Med.,
    30: 135-176,  1979.

29. Armitage,  P.  and Doll,  R. The  age distribution  of cancer and
    a multistage theory of carcinogenesis. Br. J. Cancer, 8:  1-12,
    1954.

30. Cairns,  J. Mutation and selection and the natural history  of
    cancer.  Nature,  225: 197-200,  1975.

31. Ames,   B.N.,   Durston,  W.E.,  Yamasaki,  E., and Lee,  F.D.
     Carcinogens are mutagens: A simple test system combining liver
     homogenates for  activation  and bacteria for detection. Proc
     Natl Acad Sci USA,  70: 2281-2285,  1973.

 32. Maher,  V.M.  and McCormick,  J.J.  Effect of DNA repair on the
      cytotoxicity  and  mutagenicity  of  UV  irradiation  and  of
      chemical  carcinogens  in normal  and  xeroderma pigmentosum
      cells.  In: J.M. Yuhas,  R.W.  Tennant and J.D. Regan  (eds.),
                                25

-------
    Biology  of  Radiation Carcinogenesis, pp. 129-145,  New York:
    Raven Press.  1976.

33. Kraemer,  K.H.  Oculo-cutaneous  and  internal  neoplasms  in
    xeroderma   pigmentosum:   Implications   for   theories   of
    carcinogenesis.  In:  B.  Pullman, P.O.P,  Ts'o and H.  Gelboin
    (eds.),    Carcinogenesis:   Fundamental    Mechanisms    and
    Environmental  Effects,   pp.  503-507,  New  York:  D.   Reidel
    Publishing Co. 1980.

34. Glover,  T.W., Chang,  C.C.,  Trosko, J.E.,  and  Li,   S.S.L.
    Ultraviolet    light    induction    of    diphtheria    toxin
    resistant-mutants in  normal  and xeroderma pigmentosum human
    fibroblasts. Proc Natl Acad Sci USA, 76: 3982-3986, 1979.

35. Cleaver,  J.E.  and  Trosko,  J.E.   Absence   of   excision of
    ultraviolet-induced   cyclobutane    dimers    in   xeroderma
    pigmentosum. Photochem.  Photobiol.,  11: 547-550, 1970.

36. Lieberman, M.W.,  Baney,  R.N., Lee, R.E.,  Sell,  S., and Farber,
    E. Studies on DNA. repair in human  lymphocytes  treated with
    ultimate carcinogens and alkylating agents.  Cancer Research,
    3: 1292-1306,  1971.

37. Weinberg, R.A. The action of oncogenes  in the cytoplasm and
    nucleus. Science, 230: 770-776, 1985.
                               26

-------
38. Bishop, J.M. Cellular  oncogenes  and retroviruses.  Ann.  Rev.



    Biochem.,  52: 301-354, 1983.
39. Weinberg,  R.A.   Tumor  suppressor   genes.   Science,   254:



    1138-1146, 1991.







40. Knudson, A.G. Hereditary cancer, oncogenes, and antioncogenes.



    Cancer Research,  45:  1437-1443, 1985.







41. Pitot,  H.C.,  Goldsworthy,  T.,  and  Moran,   S.  The natural



    history  of  carcinogenesis:    Implications  of  experimental



    carcinogenesis  in the genesis of human  cancer.  J.  Supramol.



    Struct.  Cellul.  Biochem.,  17:  133-146,  1981.







42. Tennant,  R.W.,  Margolin,  B.H.,  Shelby,  M.D.,  Zeiger,  E.,



    Haseman,  J.K.,   Spalding,  J.,  Caspary,  W.,  Resnick,  M.<



     Stasiewicz,  S.,  Anderson,  B., and  Minor, R. Prediction  of



     chemical carcinogenicity  in  rodents  from in vitro  genetic



     toxicity assays. Science,  235: 933-941, 1987.







 43. Nowell,  P.C. The  clonal evolution  of tumor cell population.



     Science, 194: 23-28, 1976.







 44. Fialkow, P.J. Clonal origin and stem cell evolution of human



     tumors.  In: J.J. Mulvihill,  R.W.  Miller  and J.F. Fraumerii



      (eds.). Genetics of Human Cancer, pp. 439-453, New York: Raven



     "press.  1977.




                                27

-------
 45.  Trosko,   J.E.,  Chang,  c.C.,  and  Madhukar,  B.V.  Cell-cell
     communication:  Relationship of stem cells to the carcinogenic
     process.   In:  D.E.  Stevenson, J.A.  Popp,   J.M.  Ward,  R.M.
     McClain,   T.J.  siaga  and  H.C.  Pitot   (eds.),  Mouse  Liver
     Carcinogenesis:   Mechanisms  and  Species  Comparisons,   pp.
     259-276,  New York:  Alan R.  Liss,  Inc..  1990.

 46.  Klein-Szanto, A.J.P.  and Slaga,  T.J. Effects of peroxides on
     rodent  skin: Epidermal  hyperplasia and tumor promotion.  J.
     Invest. Dermatol.,. 79:  30-34,  1982.

 47.  Argyris,  T.S.  Tumor  promotion  by  regenerative   epidermal
     hyperplasia  in  mouse skin. J.   Cutaneous Pathol.,  9: 1-18,
     1982.

 48.  Frei, J.v. and  Stephens,  P. The correlation of promotion  of
     tumour growth and of induction  of hyperplasia in epidermal
     two-stage Carcinogenesis. Br.  J. Cancer, 22: 83-92,  1968.

49. Trosko,  J.E., Chang, C.C., and Medcalf, A. Mechanisms of tumor
    promotion:  Potential  role  of  intercellular  communication.
    Cancer Invest, 1:  511-526, 1983.
50. Trosko,  J.E.  and  Chang,  C.C.  Nongenotoxic  mechanisms  in
    Carcinogenesis: Role of inhibited intercellular communication.
    In:  R.w.  Hart  and F.G.  Hoerger (eds.), Banbury Report  31:
    Carcinogen Risk Assessment: New Directions in the Qualitative
                              28

-------
    and Quantitative Aspects, pp. 139-170,  Cold Spring Harbor, NY:
    Cold  Spring Harbor Laboratory Press. 1988.

51. Trosko,  J.E.  and Chang,  C.C.  Gap  junctional intercellular
    communication  in  neoplasia:  Implications  for  the  cause and
    treatment  of   cancer.   In:  L.  Herrera  (ed.),   Familial
    Adenomatous Polyposis, pp.  2.79-288, New York: Alan R. Liss,
    Inc.. 1990.

52. Trosko,    J.E.   and   Chang,   C.C.  Stem  cell   theory  of
    carcinogenesis. Toxicol. Lett., 49: 283-295,  1989.

53. Potter,  V.R. Use of two sequential applications of  initiators
    in the production of  hepatomas  in the rat: An examination of
    the  Solt-Farber  Protocol.  Cancer  Research,  44:  2733-2736,
    1984.

54. Hennings,  H.,  Shores, R., Wenk, M.L.,  Spangler, E.F.,  Tarone,
    R.,  and Yuspa,  S.H. Malignant conversion of mouse skin tumors
    is  increased  by  tumor  initiators  and  unaffected by  tumor
    promoters. Nature,  304:  67-69,  1983.

 55. Reddy,   A.L.   and   Fialkow,   P.J.   Papillomas   induced  by
    initiation-promotion differ from those induced by carcinogen
    alone.  Nature, 304: 69-71, 1983.
                                29

-------
 56. O'Connell,  J.F.,  Klein-Szanto,  A.J.P.,   Digiovanni,   D.M. ,
     Fries, J.M., and Slaga,  T.J. Malignant progression o£  mouse
     skin papillomas treated with ethylnitrosourea, NMNG, or TPA.
     Cancer Lett, 30: 269-274, 1986.

 57. Taguchi,   T.,  Yokoyama,  M.,  and  Kitamura,   Y.  Intraclonal
     conversion  from papilloma   to  carcinonma  in the  skin  of
     PgK-l'/PgK-l6 mice treated by a complete carcinogenesis process
     or by an  initiation-promotion  regimen.  Cancer Research, 44:
     3779-3782, 1984.

 58.  Trosko, J.E., Madhukar,  B.V., and Chang, C.C.  Endogenous and
     exogenous   modulation   of   gap  junctional   intercellular
     communication: Toxicological and pharmacological implications.
     Life  Sciences, 1993  (in press).

 59.  Levine, E.M., Becker, Y. ,  Boone,  C.W., and Eagle, H. Contact
     inhibition,  macromolecular synthesis, and polyribosomes in
     cultured human diploid  fibroblasts.  Proc Natl Acad Sci  USA,
     53: 350-355, 1965.
                                  «
60. Abercrombie, M.  Contact inhibition  and  malignancy.  Nature,
    281: 259-262, 1979.

61. Borek, C.  and Sachs,  L.  The difference in contact inhibition
    of cell replication between normal cells and cells transformed
                               30

-------
    by  different   carcinogens.  Proc  Natl  Acad  Sci  USA,   55:
    1705-1711,  1966.

62. Kanno,  Y. Modulation of cell communication and carcinogenesis.
    Jpn J Physiol, 35: 693-707,  1985.

63. Loewenstein, W.R. Permeability of membrane  junctions. Ann N Y
    Acad Sci,  137: 441-472, 1966.

64. Fentman,    I.S.,    Hurst,    J-',     Ceriani,     R.L.,    and
    Taylor-Papadimitriou,    J.    Junctional    intercellular
    communication pattern  of cultured human breast cancer cells.
    Cancer Research,  39: 4739-4743, 1979.

65. Yamasaki,  H.,  Hollstein,  M., Mesnil,  M., Martel,  N.,  and
    Aguelon,  A.M. Selective lack of intercellular communication
    between transformed  and nontransformed  cells  as  a common
    property  of chemical and oncogene transformation of BALB/c3T3
    cells.  Cancer Research,  47:  5658-5664,  1987.

 66. Saez,  J.C.,  Spray, D.C.% and Hertzberg, E.L.  Gap junctions:.
     Biochemical  properties  and functional   regulation  under
     physiological and toxicological conditions. In Vitro Toxicol.,
     3: 69-86, 1990.
                                31

-------
 61. Pound, A.W. and McGuire,  L.J.  Repeated partial hepatectomy as
     a promoting  stimulus for carcinogenic  response  of liver to
   .  nitrosamines in rats. Br. J. Cancer,  37: 585-594,  1978.

 68. Yancey, S.B., Easter, D., and Revel, J.P. Cytological changes
     in gap junctions during  liver  regeneration.  J.  Ultrastruct.
     Res.,  67-.  229-242,  1979.

 69. Massey,  K.D., Minnich, B.N.,  and Burt, J.M.  Arachidonic acid
     and  lipoxygenase metabolites  uncouple neonatal  rat  cardiac
     myocyte pairs.  Am J Physiol,  263:  C494-C501,  1992.

 70.  Fluri,  G.S.,  Rudisuli, A., Willi, M., Rohr,  S.,  andWeingart,
     R.  Effects  of  Arachidonic  Acid  on  the  Gap  Junctions of
     Neonatal Rat  Heart Cells. Pflugers Arch, 417:  149-156,  1990.

 71.  Trosko,  J.E., Madhukar,  B.V.,  Hasler, C.,  and  Chang,  C.C.
     Modulated   intercellular  communication:   Consequences  of
     extracellular    molecules     triggering    intracellular
     communication.  In:  K.V. Honn,  L.J.  Marnett,  S.  Nigam and T.
    Walden  (eds.),   Eicosanoids  and Other Bioactive  Lipids in
    Cancer and Radiation Injury, pp. 285-295, Boston: Kluwer Acad.
    Publ. 1989.
72. Agarwal, R.  and Daniel, E.E. Control of gap junction formation
    in  canine  trachea  by arachidonic acid metabolites.  Am  J
    Physiol, 250: 495-505, 1986.
                               32

-------
73. Trosko,  J.E.,  Jone,  C.,  and  Chang,  C.C. The  role of tumor



    promoters on phenotypic alterations  affecting intercellular



    communication  and tumorigenesis.  Ann  N Y  Acad  Sci,  407:



    316-327, 1983.







74. Chen, T.H.,  Kavanagh,  T.J.,  Chang,  C.C., and Trosko,  J.E.



    Inhibition of  metabolic  cooperation  in Chinese hamster V79



    cells by various organic  solvents and simple compounds. Cell



    Biol Toxicol,  1: 155-171,  1984.







75. Columbano, A., Ledda-Columbano, G.M., Ennas,  M.G.,  Curto, M.,



    Chelo,  A., and Pani, P.  Cell proliferation and  promotion  of



    rat   liver   carcinogenesis:  Different   effect  of   hepatic



    regeneration   and   mitogen  induced   hyperplasia  on   the



    development   of   enzyme-altered  foci.   Carcinogenesis,   11:



    771-776,  1990.







 76.  Parkinson,     E.K.     and    Emmerson,    A.     Non-promoting



    hyperplasiogenic agents do not mimic the effects of phorbol,



     12-myristate,  13-acetate on terminal differentiation of normal



     and  transformed  human  keratinocytes.  Carcinogenesis,   5:



     687-690, 1984.
 77. Sisskin, E.E.,  Gray,  T., and Barrett, J.C.  Correlation between



     sensitivity  to  tumor  promotion   and  sustained  epidermal



     hyperplasia  of mice  and rats.  Carcinogenesis,  3: 403-407,



     1982.




                                33

-------
 78.  Rodrigueztarduchy,   G.  and  Lopezrivas,  A.  Phorbol  Esters
     Inhibit Apoptosis in IL-2-Dependent Lymphocytes-T. BIOC. BIOP.
     R.,  164:  1069-1075,  1989.

 79.  Moolgavkar, S.H. Multistage models for cancer risk assessment.
     In:  C.C.  Travis  fed.),  Biologically Based Methods for Cancer
     Risk Assessment, pp.  9-20, New York:  Plenum Press.  1989.

 80.  Atkinson,  M.M.,  Menko,  A.S.,  Johnson, R.G., Sheppard,  I.R.,
     and   Sheridan,   J.D.  Rapid  and   reversible  reduction   of
     junctional permeability in cells  infected with a  temperature
     sensitive mutant of  avian  sarcoma virus.  J Cell  Biol,   9:
     573-578,  1981.

 81.  Atkinson,   M.M.,   Anderson,    S.K.,   and   Sheridan,   J.D.
     Modification  of gap junctions  in  cells transformed  by a
     temperature-sensitive mutant of Rous sarcoma virus. J Membr
     Biol, 91:  53-64, 1986.

 82. Azarnia, R. and Loewenstein, W.R.  Intercellular communication
    and   the   control   of  growth:   Alteration  of   junctional
    permeability   by    the  src    gene   -   A   study   with
    temperature-sensitive mutant Rous  sarcoma virus. J  Membr Biol,
     82: 191-205,  1984.

83. Azarnia,  R., Reddy,  S., Kimiecki,  T.E.,  Shalloway,  D.,  and
    Loewenstein,  W.R.  The  cellular src  gene product regulates
                              34

-------
    junctional cell-to-cell communication. Science, 239: 398-401,
    1988.

84. Azarnia,  R.  and  Loewenstein,  W.R.   Polyomavirus middle  T
    antigen deregulates junctional cell to cell communication.
    Mol Cell Biol,  7:  946-950, 1987.

85. Chang,  C.C.,  Trosko,  J.E..  Kung,  H.J.,  Bombick,  D.,  and
    Matsumura,   F.  Potential  role  of  the  src gene  product in
    inhibition of gap junctional  communication in NIH 3T3  cells.
    Proc Natl Acad Sci USA,  82:  5360-5364,  1985.

 86. Dotto,  G.P.,  El-Fouly,  M.H., Nelson,  C.,  and Trosko,  J.E.
    Similar  and   synergistic   inhibition  of   gap-junctional
    communication  by   ras   transformation  and  tumor  promoter
    treatment   of  mouse  primary  keratinocytes.   Oncogene,  4:
     637-641,  1989.

 87. Bignami,  M.,  Rosa,  S.,  Falcone, G.,  Tato, F., Katoh, F., and
     Yamasaki,   H.  Specific  viral  oncogenes  cause   differential
     effects  on  cell-to-cell  communication  relevant  to  the
     suppression of  the  transformed phenotype by  normal cells.
     Molec. Carcinogenesis,  1:  67-75, 1988.

  88. El-Fouly,  M.H., Trosko, J.E., Chang, C.C., and  Warren,  S.T.
     Potential  role of the human  Ha-ras oncogene in the inhibition
                                 35

-------
    of gap junctional  intercellular communication.  Mol  Carcinog,
    2: 131-135, 1989.

89. El-Fouly,  M.H.,  Trosko,  J.E.,  and  Chang,  C.C. Phenotypic
    transformation and inhibition of gap-junctional  intercellular
    communication in epithelial and mesenchymal  cells by the neu
    oncogene.  4th  Annual Oncogene  Meeting,  Frederick,  MD: July
    5-9,  1988.(Abstract)

90. Kalimi, G.H., Hampton, L.L., Trosko, J.E., Thorgeirsson, S.S.,
    and Huggett, A.C. Homologous and heterologous gap-junctional
    intercellular   communication    in    v-raf-,   v-myc-,   and
    v-raf/v-myc-transducted rat liver epithelial cell lines. Mol
    Carcinog, 5: 301-310, 1992.

91. de Feijter, A.W. ,• 'Ray, J.S.,  Weghorst, C.M., Klaunig,  J.E.,
    Goodman,   J.I.,  Chang, C.C.,  Ruch,  R.J.,  and  Trosko,  J.E.
    Infection  of   rat   liver   epithelial  cells  with  v-Ha-ras:
    Correlation  between  oncogene   expression;  gap  junctional
    communication,  and  tumorigenicity.  Mol  Carcinog, 3:  54-67,
    1990.

92. de Feijter, A.W., Trosko,  J.E., Krizman, D.B., Lebovitz, R.M.,
    and Lieberman, M.W.  Correlation of increased levels of Ha-ras
    T24 protein with extent of loss of  gap junction function  in
    rat liver .epithelial cells. Mol Carcinog, 5:  205-212,  1992.
                               36

-------
93. Kalimi, G., Chang, C.C.,  Edwards,  P., Dupont, E., Madhukar,
    B.V., Stanbridge,  E.,  and Trosko, J.E. Re-establishment of gap
    junctional  communication  in a  non-tumorigenic  Hela-normal
    human  fibroblast  hybrid.   Proc.  Am.  Assoc.  Cancer Res., 31:
    319, 1990.

94. Lee, S.W., Tomasetto, C.,  and Sager, R. Positive selection  of
    candidate tumor-suppressor genes by subtractive hybridization.
    Proc.  Natl. Acad.  Sci.  USA,  88:  2825-2829,  1991.

95. Mehta, P.P., Bertram, J.S.,  and  Loewenstein, W.R. The actions
    of retinoids  on cellular growth correlate with their actions
    on  gap  junctional  communication.   J.   CELL  BIOL.,  108:
     1053-1065,  1989.

 96. Mehta,,  P.P.,  Bertram,  J.S.,  and  Loewenstein,  W.R.  Growth
     inhibition of transformed cells  correlates with the junctional
     communication with normal cells. Cell, 44: 187-196, 1986.

 97. Rogers, M., Berestecky, J.M., Hossain, M.Z., Quo,  H.M., Kadle,
     R., Nicholson, B.J.,and  Bertram,  J.S. Retinoid-Enhanced  Gap
     Junctional Communication Is Achieved by Increased  Levels  of
     Connexin-43  Messenger RNA  and  Protein. Mol.  Carcinogen.,  3:
     335-343,  1990.

  98. Brummer,  F.,  Zempel, G.,  Buhle, P.,  Stein, J.C., and Hulser,
     D.F.  Retinoic acid modulates  gap  junctional permeability: A
                                 37

-------
      comparative study  of dye  spreading  and ionic  coupling  in

      cultured cells.  Exp.  Cell Res.,  196: 158-163, 1991.




  99.  Rivedal,  E.  and  Sanner,  T.  Regulation  of gap  junctional

      communication  in Syrian hamster embryo cells by retinoic acid

      and 12-O-tetradecanoyl-phorbol-13-acetate. Carcinogenesis, 13:

      199-203,  1992.
               »



 100.  Azarnia,  R.  and  Russell,  T.R. Cyclic AMP effects  on  cell  to

      cell   junctional  membrane  permeability during   adipocyte

      differentiation  of 3T3-L1 fibroblasts.  J. CELL BIOL.,  100:

      265-269,  1985.




 101.  Flagg-Newton,  J.L.,  Dahl,  G.,  and Loewenstein,  W.R.  Cell

      junction  and cyclic AMP: Upregulation  of  junctional membrane

     permeability   and'   junctional    membrane   particles   by

     administration  of  cyclic nucleotide  or phosphodiesterase

     inhibitor. J. Mernb. Biol., 63: 105-121, 1981.




102. Demaziere,  A.M.G.L.  and  Scheuerman,   D.W.  Increased  gap

     junctional  area  in  the  rat  liver after administration of

     dibutyrl cAMP.  Cell Tiss Res, 239: 651-655, 1985.




103. Veld,   P.I.,  Schuit,  F.,   and Pipeleers,  D. Gap  junctions

     between pancreatic B-cells are modulated  by cyclic AMP. Eur.

     J. Cell Biol.,  36: 269-276, 1985.
                                38

-------
104.  Zhang,  L.X.,  Cooney,  R.V.,  and  Bertram,  J.S.  Carotenoids
     enhance  gap  junctional  communication  and  inhibit  lipid
     peroxidation in C3H/10T1/2 cells: Relationship to their cancer
     chemopreventive action. Carcinogenesis,  12: 2109-2114, 1991.

105. Ruch, R.J., Madhukar,  B.V.,  Trosko,  J.E.,  and Klaunig,  J.E.
     Reversal   of   ras-induced   inhibition   of   gap   junctional
     intercellular communication, transformation and tumorigenesis
     by lovastatin. Molec. Carcinogenesis, 1992.(in press)

106. Fishman, G.I., Moreno,  A.P.,  Spray,  D.C., and Leinwand, L.A.
     Functional analysis of  human  cardiac  gap  junction  channel
     mutants. Proc. Natl, Acad.  Sci. USA,  88: 3525-3529,  1991.

 107.  Zhu,   D.,   Caveney,   S.,   Kidder,   G.M., and Naus,  C.C.G.
      Transfection of  C6 '  glioma cells  with  Connexin-43  cDNA:
      Analysis  of  expression,  intercellular  coupling,  and  cell
      proliferation.  Proc.  Natl. Acad.  Sci. USA, 88:  1883-1887,
      1991.
                       •i

 108. Eghbali, B., Kessler,  J.A., and Spray,  D.C. Expression of gap
     ' junction  channels  in  communication-incompetent  cells  after
      stable  transfection with cDNA encoding connexin-32. P. NAS.
      US.,  87:  1328-1331, 1990.
                                 39

-------
 109.  Potter,  V.R.  Phenotypic diversity in experimental hepatomas:
      The concept of partially blocked ontogeny.  Br. J. Cancer, 35:
      1-23,  1978.

 110.  Pierce,  G.B. Neoplasms, differentiation and mutations. Am. J.
      Pathol.,  77:  103-118,  1974.

 111.  Markert,  C.  Neoplasia:  A disease of cell differentiation.
      Cancer Research,  28i  1908-1914,  1968.'

 112.  Nakano,  S.,  Ueo,  H.,  Bruce,  S.A.,   and  Ts'o,  P.O.P.  A
      contact-insensitive  subpopulation  in  Syrian  hamster  cell
      cultures with a -greater susceptibility to chemically induced
      neoplastic  transformation.  Proc  Natl  Acad  Sci  USA,   82:
      5005-5009, 1985.

113.  Lo, C.W.  Communication compartmentation and pattern  formation
      in development. In: M.V.L.  Bennett and  D.C.  Spray (eds.)-,  Gap
     Junctions, pp.  251-264,  Cold Spring Harbor,  New York: Cold'
     Spring Harbor Laboratory.  1985.

114. Chang,  C.C.,  Trosko, J.E.,  El-Fouly, M.H., Gibson-D'Ambrosio,
     R.E.,  and  D'Ambrosio,  S.M.  Contact  insensitivity  of  a
     subpopulation of  normal human  fetal  kidney epithelial cells
     and  of  human  carcinoma cell  lines.  Cancer  Research,   47:
     1634-1645, 1987.
                                40

-------
115. Chang,  C.C.,  Nakatsuka,  S.,  Kalimi,  G., Trosko,  J.E.,  and



     Welsch,  C.W.  Characterization of two types  of  normal human



     breast  epithelial  cells   that   are  either  deficient  or



     proficient in gap  junctional intercellular communication. J



     Cell Biochem Suppl, 14B: 331, 1990.







116. Keski, J.  and Moses, H.C. Growth inhibitory polypeptide in the



     regulation of cell  proliferation. Med. Biol.,  65:  13-20, 1987.







117. Sonnenschein, C.,  Olea, N.,  Pasanen,  M.E.,   and Soto, A.M.



     Negative controls of cell proliferation: Human prostate cancer



     cells and androgens. Cancer Research, 49: 3474-3481,  1989.







118. Kimchi, A.,  Wang,  X.-F., Weinberg,  R.A.,  Cheifetz,  S.,  and



     Massague,  J.  Absence  of   TGF-beta receptors   and   growth



     inhibitory  responses'  in retinoblastoma cells. Science, 240:



     196-198,  1988.







119. de Rooij, D.G., Lok, D., and Weenk, D.  Feedback regulation of



     the proliferation of the undifferentiated spermatogonia in the



     Chinese  hamster by  the differentiating spermatogonia. Cell-



     Tissue Kinet,  18:  71-81,  1985.
 120.  Zhu,  D.G., Kidder, G.M., Caveney, S., and Naus, C.C.G. Growth



      Retardation    in  Glioma    Cells   Cocultured   with   Cells



      Overexpressing a Gap Junction Protein. Proc Natl Acad Sci USA,



      89: 10218-10221, 1992.




                                 41

-------
 121.  Paget,  s.  The distribution of  secondary  growth in cancer of
      the  breast.  Lancet, '!•  571-573, 1889.

 122.  Tokunaga, M. ,  Land, C.E.,  and Tokuoka,  S.  Follow-up studies
      of breast  cancer incidence among  atomic bomb  survivors.   j
      Radiat  Res  (Suppl.) 32:201-211,  1991.

 123.  Trosko, J.E., Chang, C.C., Madhukar,  B.V.,  and  Klaunig,  J.E.
      Chemical,  oncogene  and  growth  factor  inhibition  of   gap
      junction   intercellular  communication:     An   integrative
      hypothesis of carcinogenesis.  Pathobiol.,  58:  265-278,  1990.

 124. Verma, A.K.  and Boutwell, R.K. Effects of dose and duration of
     treatment with the tumor promotiong agent, TPA,  on mouse  skin
     carcinogenesis. Carcinogenesis, 1: 271-276, 1986.

 125. Goldsworthy, T.,  Campbell,  H.,  and  Pi tot,  H.C. The natural
     history and dose-response  characteristics of enzyme-altered
     foci     in    rat   liver    following   phenobarbital    and
     diethylnitrosamine administration.  Carcinogenesis, 5: 67-71,
     1984.

126. Demi, E.  and Oesterle,  D. Dose response  of  promotion by
     polychlorinated biphenyls and chloroform in  rat  liver foci
     bioassay. Arch Toxicol,  60-.  209-211,  1987.
                               42

-------
127. Maekawa, A.,  Onodera,  H.,   Ogasawara,  H.,  Matsushima,  Y.,
     Mitsumori,  K.,' and Hayashi,  Y.  Threshold dose dependence in
     phenobarbital promotion of rat hepatocarcinogenesis initiated
     by diethylnitrosamine. Carcinogenesis,  13: 501-503, 1992.

128. Pereira,   M.A.,   Herren-Freund,    S.L.,   and   Long,   R.E.
     Dose-response  relationship  of  phenobarbital  promotion  of
     diethylnitrosamine  initiated  tumors  in  rat  liver.' CANCER
     LETT.,  32: 305-311, 1986.

129. Yancey, S.B., Biswal,  S.,  and Revel,  J.P. Spatial and temporal
     patterns   of  distribution  of   the  gap   junction   protein
     connexin43   during  mouse  gastrulation  and  organogenesis.
                                    «
     Development,  114:  203-212,  1992.

 130.  Traub,  O.,  Druge,  P.M.,  and Willecke,  K.  Degradation  and
      resynthesis  of gap junction protein  in plasma membranes of
      regenerating liver after partial hepatectoray'or cholestasis.
      Proc Natl Acad Sci USA, 80: 755-759,  1983.

 131. Bryant,   P.J.   and   Fraser,  S.E.  Wound   healing,   cell
      communication,  and  DNA  synthesis  during  imaginal  disc
      regeneration in Drosophila. Dev  Biol,  127:  197-208, 1988.

 132. Trosko,  J.E.,  Chang,  C.C.,  and Netzloff,  M. The  role  of
      inhibited    cell-cell    communication    in   teratogenesis.
      Teratogenesis Carcinog Mutagen,  2:  31-45,  1982.
                                 43

-------
 133.  Warner,  A.E.,  Guthrie, S.C., and Gilula,  N.B.  Antibodies to
      gap   functional   protein  selectively   disrupt   junctional
      communication  in  the early  amphibian  embryo. Nature,  311:
      127-131,  1984.

 134.  Trosko,  J.E.,  Jone,  C.,  and Chang,  C.C.  Inhibition of  gap
      junctional-mediated intercellular  communication in  vitro  by
      Aldrin, Dieldrin, and Toxaphene: A possible cellular mechanism
      for  their  turnor-promotiong  and   neurotoxic  effects.  Mol
      Toxicol,  1: 83-93,  1987.

 135.  Ye, Y.X., Bombick,  D., Hirst, K.,  Zhang, G.X., Chang, C.C,,
      Trosko, J.E., and Akera,  T. The modulation of gap  junctional
      communication by gossypol in  various mammalian  cell  lines  in
      vitro. Fundam. Appl.  Toxicol., 14:  817-832,  1990.

136.  Saffitz,  J.E.,  Hoyt,  R.H.,   Luke,   R.A.,  Kanter,  H.L.,  and
      Beyer,  E.G.  Cardiac myocyte interconnections  at  gap junctions
      - Role in normal  and abnormal electrical conduction. Trend.
     Cardiovasc Med,  2:  56-60,  1992.

137. Kleber,  A.G.,  Riegger,   C.B.,  and  Janse,   M.J.   Electrical
     uncoupling and  increase  of  extracellular resistance after
     induction of  ischemia in isolated, arterially perfused rabbit
     papillary muscle.  Circ Res, 61:  271-279, 1987.
                                44

-------
138. Tanaka,   T.,  Sakai,   M.,   Fujimoto,  K.,  and   Ogawa,   K.
    ' Morphometric analysis of gap junctions in  the rat lens during
     cataract  formation.  Acta  Histochem Cytochem,  23:  781-792,
     1990.

139. Smith,  J.H.,  Green,  C.R.,  Peters,  N.S.,  Rothery,  S.,  and
     Severs, N.J. Altered patterns of gap junction distribution in
     ischemic heart disease - An immunohistochemical study of human
     myocardium  using  laser scanning confocal microscopy. Am. J. .
     Pathol.,  139: 801-821,  1991.

140. Schellens,  J.P.,  Blange, T.,  and  de Groot,  K. Gap  junction
     ultrastructure  in rat  liver  parenchymal cells after  invivo
     ischemia. Virchows.  Arch.  [B],  53: 347-352, 1987.

141. Witkop,  C.J., White,  J.G., King,  R.A.,  Dahl, M.V.,  Young,
     W.G.,  and Sauk,  J.J.  Hereditary mucoepithelial dysplasia:  A
     disease apparently of  desmospme and gap junction formation.
     Am.  J Hum.  Genet.,  31: 414-427, 1979.

 142. Campos  de  Carvalho,   A.C.,   Tanowitz,   H.B.,  Wittner,  M.,
      Dermietzel, R.,  Roy, C., Hertzberg, E.L., and  Spray, D.C. Gap
      junction  distribution is  altered between cardiac  myocytes
      infected with Trypanosoma  cruzi . Circ Res, 70:  733-742, 1992.
                                 45

-------
BIOGRAPHICAL SKETCH:  WILLIAM H. van der SCHALIE, Ph.D.
Dr. van der Schalie received his undergraduate degree in Biology from the Michigan
State University in  1973.  His Ph.D. in Zoology was received from the  Virginia
Polytechnic Institute and State University in 1977.

Dr. van der Schalie is currently a Biologist with the Risk Assessment Forum of the
United States Environmental  Protection Agency  in Washington, DC.  As  science
coordinator for the Risk Assessment Forum, he facilitates the activities of senior EPA
scientists.  These activities are ultimately directed toward developing EPA's first
Agency-wide guidelines for ecological risk assessment.  He has been in this position
since August of 1990.  Prior to joining  EPA, Dr. van der Schalie was a Biological
Sciences Administrator and a Branch Chief with the Research Methods Branch, Health
Effects Research Division,  of the U. S. Army  Biomedical Research and Development
Laboratory.

Dr. van der Schalie is an Associate Professor of Biology (part-time) at Hood College
in  Frederick,  MD.  His past experience includes positions as a Research  Aquatic
Biologist and  a National Research Council Resident Research Associate at the U. S.
Army Biomedical  Research and Development Laboratory.  He was also a Teaching
Fellow at the University of Michigan Biological Station in Pellston, Ml.

-------
                                    Annual Review

                    A FRAMEWORK FOR ECOLOGICAL
                       RISK ASSESSMENT AT THE EPA

              SUSAN B. NoRTON.t DONALD J. RQDIER,! JOHN H. GENTILE,}
                   WILLIAM H. VAN DER SCHALIE,! WILLIAM P.  WOOD!
                                and  MICHAEL W. SLIMAK*!
                   fU.S. Environmental Protection Agency, Washington,  DC 20460
          JU.S. Environmental Protection Agency Laboratory Narragansett, Rhode Island 02592

                         (Received 9 January 1992; Accepted 13 July 1992)

      Abstract-Ecological risk assessments evaluate the likelihood of adverse ecological effects caused
      bv s Ss ors relatld to human activities such as draining of wetlands or release of chemicals The
      ?erm /rrSor i used to describe any chemical, physical, or biological entity that can induce adverse
      eff«ts on ecological components (i.e., individuals, populations, commumt.es, or ecosystems). In
      * s relSvart de a historical perspective on ecological riskassessment activities at theUS-Env,-
       onmental Protection Agency (EPA) is followed by a discussion of the EPA s "*™^ **S
      which describes the basic elements for conducting an ecological risk assessment. The Framework
      Renort" is neither a procedural guide nor a regulatory requirement within the EPA. Rather, it is
      Eded to foster'Iconsistent approach to ecological risk assessments withm the Agency, identify
      key issues, and define terminology.

      Keywords-Risk assessment    Ecological stressors
               INTRODUCTION
   Environmental problems are often complex,
with multiple causes and diverse ecological effects.
Examples include the effects of global climate
change, habitat loss, acid deposition, and multiple
chemicals present in the environment. Dealing with
such problems requires a flexible decision-making
process that can accommodate this diversity while
providing some measure of the uncertainty associ-
ated with decisions that are made. At the U.S. En-
vironmental Protection  Agency (EPA), there is
increasing interest in  using ecological risk assess-
ments as a basis for environmental decisions.
   This article examines the EPA's past, present,
and possible future utilization of ecological risk
assessment approaches. The core of the discussion
is a summary of the EPA's recently published
"Framework Report," which describes the basic
 elements of, or a framework for, ecological risk
 assessment and has been proposed as a basis for
 conducting ecological risk assessment within the
 EPA [1].                                .
    The "Framework Report" does not contain sub-
 stantive guidance on factors that are integral to the
 risk assessment process, such as analytical meth-
 ods, techniques for analyzing and interpreting data,

     •To whom correspondence may be addressed.  -
or guidance on factors influencing policy. Such is-
sues are reserved for future guidance, and plans for
developing such guidance, based on the approach
described in  the "Framework Report," are de-
scribed in the last section of this article. We hope
that this  article will broaden the audience for the
"Framework  Report" and help stimulate discus-
sions of the many issues that have been highlighted
throughout the guideline development process.

 The nature of ecological risk assessment
    Ecological risk assessments evaluate the likeli-
 hood that adverse ecological effects will occur as
 a result of exposure to stressors related to human.
 activities, such as draining of wetlands or release of
 chemicals. The term stressor is used here to describe
 any chemical, physical, or biological entity that can
 induce adverse effects on ecological components,
 that is, individuals, populations, communities, or
 ecosystems. Adverse ecological effects encompass
 a wide range of disturbances ranging from mortal-
 ity in an individual organism to a loss in ecosystem
 function. Thus, the ecological risk assessment pro-
 cess must be flexible while  providing a logical and
 scientific structure to accommodate a broad array
 of stressors  and ecological components.
    Ecological risk may be expressed in a variety of
 ways. Whereas some ecological risk  assessments
                                               1663

-------
 1664
                                        S.B. NORTON ET AL.
 may provide true probabilistic estimates of risk,
 others may be deterministic or even qualitative in
 nature. In these cases, the likelihood of adverse ef-
 fects is expressed through a. semiquantitative  or
 qualitative comparison of effects and exposure.
    The appropriate application of risk assessment
 helps meet the EPA's goal of targeting environmen-
 tal protection resources at the problems and the
 geographic areas posing the greatest risks. The fol-
 lowing section briefly discusses the  use and devel-
 opment of ecological risk assessment approaches in
 EPA programs.

 Ecological risk assessment at EPA
    Since its establishment in 1970, the EPA has at-
 tempted to protect both human health and ecolog-
 ical resources. Typically, however, activities most
 closely related to human health have received the
 highest priority. Nonetheless, assessment of risk  to
 ecological resources has been an important activ-
 ity for many programs at the EPA.  The recent re-
 port by the Science Advisory Board [2] strongly
 supported an increased emphasis on risk-based de-
 cision making and ecological risk assessment. The
 board's recommendations included that (a) the
 EPA should target its  environmental protection ef-
 forts on the basis of opportunities for the greatest
 risk reduction; (b)  the EPA should attach as much
 importance to reducing ecological risk as it does  to
 reducing  human health risk; and (c)  the  EPA
 should improve the data and analytical methodol-
 ogies that support the  assessment, comparison, and
 reduction' of different environmental risks.
   The ecological  risk activities of two EPA of-
 fices—the Office  of  Pesticides and Toxic  Sub-
stances and th'e Office of Water—illustrate the use
and development of ecological risk  assessment at
the EPA. The Office of Prevention,  Pesticides and
Toxic Substances is concerned about potential im-
pacts of pesticides and toxic chemicals on organ-
isms, including aquatic and terrestrial communities.
 Its legal mandates come from the Federal Insecti-
cide, Fungicide, and Rodenticide Act (FIFRA) and
the Toxic Substances  Control Act (TSCA).
   Both programs under the Office of Prevention,
 Pesticides and Toxic Substances—the Office of
 Pesticide Programs (OPP) and the Office of Pol-
 lution Prevention and  Toxics (OPPT)—assess risks
to ecological resources by an ecotoxicological ap-
proach: Laboratory toxicity bioassays are used to
characterize ecological effects; exposure is estimated
by using either monitoring data or models; and the
risk is estimated by comparing the two with a sim-
ple quotient, risk  = exposure/effects. Although
 the use of laboratory bioassays and the quotient
 method have provided results useful for decision-
 making with a reasonable commitment of resources,
 there has been increased interest in strengthening
 the assessment of ecological effects. Advances in-
 clude incorporating toxicity. data for higher levels
 of organization (e.g., through mesocosms), using
 simulation models to project effects at lower tro-
 phic levels to higher trophic levels, and developing
 techniques to assess exposure to multiple chemicals.
    The Office  of Water is required by the Clean
 Water Act to restore and  maintain the biological
 integrity of the nation's waters and, specifically, to
 ensure the protection and propagation of a balanced
 population of fish, shellfish, and wildlife. The EPA
 also develops methods,  including biological mon-
 itoring and  assessment  methods, for establishing
 and measuring water-quality criteria. These statu-
 tory requirements have  encouraged the Office of
 Water to  develop innovative approaches to eco-
 logical assessment. The Water Quality Act of 1987
 (Public Law 100-4) amends the decade-old Clean
 Water Act and redirects its focus from the technol-
 ogy approach, based on end-of-pipe standards, to
 full-scale implementation of the water-quality ap-
 proach, based on ambient receiving water standards.
    State water-quality standards and designated
 uses form the backbone of the water quality-based
 approach, and  the EPA criteria are developed as
 national recommendations to assist states in devel-
 oping their standards. The most commonly used
 risk-based approaches to the evaluation of water
 quality are the application of chemical-specific wa-
 ter-quality criteria and whole-effluent toxicity cri-
 teria. Both criteria have three components, the first
 characterizing the ecological  effects and the latter
 two assessing the exposure:

 1. Magnitude—what concentration of a pollutant
   (or  a pollutant parameter such  as toxicity)  is
   allowable
 2. Duration-the period of  time over which the
   predicted in-stream concentration occurs (this
   specification limits the duration of concentra-
   tion above the criteria)
 3. Frequency - how often criteria can be exceeded
   without unacceptably affecting the community.

The approaches to risk-based water-quality crite-
ria are being expanded by the Office of Water in
their development of several new types of criteria,
including chemical-specific sediment criteria and
wildlife criteria, and  biological criteria based on
community structure.

-------
                          Framework for ecological risk, assessment at EPA
                                                                                         1665
   Although the programs in the Office of Preven-
tion, Pesticides and Toxic Substances and the Of-
fice of Water have a longer history, all of the EPA's
programs are beginning to address ecological con-
cerns. Responding to the increased emphasis on
ecological issues and the need to provide for some
uniform procedures, the EPA's Risk Assessment
Council (senior managers with significant respon-
sibilities for assessment and reduction of risks) di-
rected the Risk Assessment Forum to develop
ecological risk assessment guidelines using the same
open process that was used  in developing human
health risk assessment guidelines. The first docu-
ment produced from the process, the "Framework
Report," is discussed below.

     EPA'S FRAMEWORK FOR ECOLOGICAL
               RISK ASSESSMENT
    The "Framework Report" describes a basic and
 flexible structure for conducting ecological risk as-
sessments within the EPA. The framework provides
 principles not only for estimating risks from chem-
 icals, but  also for predicting impacts  from non-
 chemical stressors (e.g., habitat loss from human
 activities), and retrospectively for assessing site-spe-
 cific impacts. The "Framework Report" is neither
 a procedural guide  nor a regulatory requirement
 within the EPA. Rather, it  is intended to foster a
 consistent approach to ecological risk assessment
 within the EPA, identify key issues,  and define
 terminology.
    The framework  is conceptually, similar to the
 approach used for human  health risk assessment
 but is distinctive in its emphasis in three areas.
 First, ecological risk  assessment can consider ef-
 fects beyond those on individuals of a single spe-
 cies and may examine a population, community, or
 ecosystem. Second, there is no single set of ecolog-
 ical values to be protected that can be generally ap-
 plied. Rather, values are selected from a number of
 possibilities based on both scientific and policy con-
 siderations. Finally, there  is an increasing aware-
  ness of the need for ecological risk assessments to
  consider nonchemical as well as chemical stressors.
     The framework for ecological risk assessment is
  illustrated in Figure 1. The risk assessment process
  is shown within the bold line. Figure 1 also illus-
  trates the interaction of risk assessment with data
  acquisition, data verification, and monitoring. la
  the "Framework Report,"  a distinction is made be-
  tween data acquisition (which is outside the risk as-
  sessment process) and data analysis  (which  is an
  inteeral part of an ecological risk assessment). The
point in the risk assessment process. At that point,
the risk assessment stops, the necessary data are ac-
quired, then the assessment resumes. Verification
and monitoring can help determine the overall ef-
fectiveness of the framework approach, provide
necessary feedback concerning the need for future
modifications of the framework, help evaluate the
effectiveness and practicality of policy decisions,
and  indicate the need for new or improved scien-
tific techniques [3].
   Finally, whereas risk assessment and risk man-
agement are distinct processes,  Figure 1 indicates
two  points of interface between these two processes
during discussions  between the risk assessor and
risk manager. At the initiation  of the risk assess-
ment, the risk manager can help ensure that the risk
assessment will provide  information relevant to
making decisions on the issues under consideration,
while the risk assessor can ensure that the risk as-
sessment addresses all relevant ecological concerns.
 Effective communication is also important at the
end of the risk assessment process to provide the
 risk manager with a full and complete understand-
 ing  of the assessment's conclusions, assumptions,
 and limitations.
    The remainder of this section discusses the three
 major phases of ecological risk assessment shown
 in Figure 1: problem formulation, analysis, and
 risk characterization. To illustrate the described
 concepts and issues, a simplistic, hypothetical ex-
 ample  concerning nutrient loads to an estuary will
 also be discussed.

 Problem formulation
     Problem formulation is a planning and scoping
 process that links the regulatory or  management
 goal to the risk assessment. Its end product is a
 conceptual model that identifies the environmen-
 tal values to be protected (the assessment end •
 points), the data  needed, and the analyses to be

     The initial steps in problem formulation include
  the identification and preliminary characterization
  of the stressor, the ecosystem potentially at risk,
  and the ecological effects. Performing this analy-
  sis is an interactive process; foe example, gathering
  information on the characteristics of a stressor
  helps to define the ecosystems potentially at risk
  from the stressor as well as the ecological effects
  that may result. The ecosystem within whiebeffects
  occur provides  the ecological context for the assess-
  ment. Knowledge of the ecosystem potentially at
   risk can help identify ecological components (i.e.,

-------
  1666
                                        S.B. NORTON ET AL.
           Discussion
          between the
         Risk Assessor
             and
         Risk Manager
           (Planning)
                               Ecological Risk Assessment
                                PROBLEM FORMULATION
                                A
                                N
                                A
                                L
                                Y
                                S
Characterization
     of
   Exposure
Characterization
     of
  Ecological
    Effects
                                 RISK CHARACTERIZATION
                                               o
                                               09
                                               5T
                                              o
                                                                                       o
                                                                                       O>
                                                                                       a

                                                                                       o
                                                                                       3
                                                                                       3
                                                                                       
-------
                          Framework for ecological risk assessment at EPA
                                         1667
assessment end point. For example, a decline in a
sport fish population (the assessment end point)
may be evaluated by using laboratory studies on
the mortality of surrogate species such as the fat-
head minnow (the measurement end point).
   Assessment and measurement end points may
involve ecological components from any level of
biological organization, ranging from individual
organism's to the ecosystem itself. In general, the
use of a suite of assessment and measurement end
points at different organizational levels  can build
greater confidence in the conclusions of the risk as-
sessment and ensure that all important end points
are  evaluated. In some situations, measurement
end points at one level of organization may be re-
lated to an assessment end point at a higher level.
 For example, measurement end points at the indi-
vidual level (e.g., mortality, reproduction, and
 growth) could be used in a model to predict .effects
 on an assessment end point at the population level
 (e.g., viability of a trout population in  a stre'am).
    Sound professional judgment is necessary for
 proper assessment and measurement end point se-
 lection, and it is important that both the selection
 rationale and the linkages  between measurement
 end points, assessment end points, and policy goals
 be clearly stated. More detailed discussions of end
 points and selection criteria can be found in Suter
 [4,5], Kelly and Harwell [6], U.S. Department of
 the Interior [7], and EPA [8],
     The initial evaluation of stressors, the ecosystem
 potentially at risk, and ecological effects are inte-
 grated with the end points to develop a conceptual
 model for the assessment. The conceptual model
  describes how the stressor might affect ecological
  components of the natural environment  [9]. For
  example, the stressor may cause adverse effects by
  interacting directly with an ecological component.
  A stressor may also cause adverse effects indirectly,
  for example, by affecting the food or habitat on
  which the ecological component of interest depends.
     The conceptual model includes a description of
  possible exposure scenarios, which are qualitative
  descriptions of how the various ecological compo-
  nents co-occur with or contact the stressor.  Each
  scenario is defined in terms of the stressor, the type
  of biological system and principal ecological com-
   ponents, how the stressor will contact or interact
   with the  system, and the  spatial and temporal
   scales. Finally the conceptual model also describes
   the approaches, analyses, and data needed to con-
   duct the assessment.
      Mthoueh there may be many ways that a stres-
considered most likely to contribute to risk are se-
lected for further evaluation in the analysis phase.
Professional judgment is needed to select the most
appropriate focus for the risk assessment, and it is
important to document the selection rationale.
   In our hypothetical  example,  a risk manager
may be concerned about possible effects of nutri-
ent inputs to an estuary. During the problem for-
mulation phase, the ways that nutrient inputs may
cause effects in the estuary are described. For ex-
ample, nutrient loads may directly alter benthic
community structure; ultimately result in decreased
dissolved oxygen levels, which then may increase
mortality rates of fish or invertebrates; or reduce
aquatic vegetation abundance, which then may in-
directly affect wildlife and fish populations that de-
pend  on the plants. Although several  of  these'
 hypotheses may be evaluated further in the analy-
 sis phase, for the purposes of this example we will
 focus on effects to the aquatic vegetation. The as-
 sessment end point may be the maintenance of the
 abundance and distribution of several species of
 aquatic vegetation as established by baseline mea-
 surements. The measurement end points could be
 growth and reproduction measurements of these
 species in the laboratory and field.
 Analysis
    The analysis phase of ecological risk assessment
 consists of the technical evaluation of the data on
 the potential effects and exposure of the stressor.
 The analysis phase is  based on the conceptual
  model developed during problem formulation. Al-
  though this phase consists of characterization of
  ecological effects and characterization of exposure,
  the dotted line in Figure 1 illustrates that the two
  are best performed interactively. An interaction be-
  tween the two elements will ensure that the charac-
  terized ecological effects are compatible with the
  biota and exposure pathways identified in the ex-
  posure characterization. The outputs of ecological
  effects characterization and exposure characteriza-
  tion are summary profiles that are used in the risk
  characterization phase.
     It is important to describe clearly and estimate
   quantitatively the assumptions and uncertainties in-
   volved in both analysis steps. In the majority of as-
   sessments, data will not be available for all aspects
   of these analyses, and  those data that are available
   may be of questionable or unknown quality. Typ-
   ically, the assessor will have to rely on a number of
   assumptions with varying degrees of uncertainty as-
   sociated with each. These assumptions will  be

-------
   1668
                                          S.B. NORTON ET AL.
  inferences based on analogy with similar chemicals
  and conditions, estimation  techniques,  and so
  forth, all of which contribute to the overall uncer-
  tainty. The uncertainties in  these  two steps are
  brought forward and summarized during  risk
  characterization.
     Characterization of exposure. The objective of
  the exposure characterization is to combine the spa-
  tial and temporal distributions of both the ecolog-
  ical component  and the  stressor to evaluate the
  co-occurrence or contact between the ecological
  component and the stressor. The way exposure is
  characterized will depend on the stressors being
  evaluated and the assessment and measurement end
  points. In the case of physical alterations of com-
  munities and ecosystems,  exposure can be broadly
  expressed as co-occurrence, for example, the co-
  occurrence of a wetland community with fill ma-
  terial. Exposure analyses of individuals often focus
  on actual contact with the stressor,  as organisms
  may not contact all of the stressor  present in an
  area. For chemical stressors, the analyses may fo-
  cus further on the amount of chemical that is bio-
  available,  that is, available  for  uptake by  the
  organism. Some  chemical exposure  analyses also
  follow the chemical within  the organism's body and
  estimate the amount that reaches the target organ.
    In order to estimate exposure appropriately, the
  temporal and spatial scale of the stressor distribu-
.  tion must be compatible with that of the ecologi-
  cal component. A temporal scale may encompass
  the life span of a species, a particular life stage, or
  a particular cycle, for example, the long-term suc-
  cession of a forest community.  A spatial scale may
 encompass a forest, a lake, a watershed, or an en-
 tire region. Stressor timing relative to organism life
 stage and activity patterns can greatly influence the
 occurrence of adverse effects. Even short-term
 events may be significant if they coincide with crit-
 ical life stages. Periods of reproductive activity may
 be especially important, because early life stages are
 often more sensitive to stressors and adults may
 also be more vulnerable at this time.
    The product of the characterization of exposure
 is an exposure profile that quantifies the magnitude
 and the spatial and temporal pattern of exposure
 for the scenarios developed during problem formu-
 lation. Exposure profiles can be expressed by using
 a variety of units.  For chemical stressors operating
 at the organism level, the usual metric is expressed
 in dose units, for example, milligrams per kilogram
 body weight per day. For higher levels of organi-
 zation, such as an entire ecosystem, exposure may
  area per time. For physical disturbance, the expo-
  sure profile may be expressed in other terms, such
  as percentage of removed habitat or the extent of
  flooding per year.
     In our nutrient loading example, exposure may
  be characterized by modeling or measuring the nu-
  trient concentrations in the  parts of the estuary
  suitable for plant growth, say, in depths of 2 m or
  less. The temporal and spatial variation would also
  be addressed, and the uncertainty in the measure-
  ments and model would be discussed.
     Characterization of ecological effects.  The re-
  lationship between the stressor and the assessment
  and measurement end points identified during
  problem formulation is analyzed in characteriza-
  tion of ecological effects. The analysis focuses on
  describing the relationship between the amount of
  stressor and the  magnitude of ecological effects
  elicited. Any extrapolations from measurement end
  points to assessment end points are also conducted
 during this phase. Finally, the  evidence for a causal
 relationship between the stressor and the measure-
 ment and assessment end points is evaluated.
    Data from both field observations  and experi-
 ments in controlled settings can be used to evalu-
 ate ecological effects. Controlled laboratory and
 field tests  (e.g., mesocosms)  can provide strong
 causal evidence linking a stressor  with a response
 and can also help discriminate between multiple
 stressors. Data from laboratory studies tend to be
 less variable than those from field studies, but be-
 cause environmental factors  are controlled, re-
 sponses may differ  from 'those  in the natural
 environment. Observational  field studies (e.g.,
 comparison to reference sites) provide environmen-
 tal realism that laboratory studies lack, although
 the presence of multiple stressors  and  other con-
 founding factors (e.g., habitat quality) in the nat-
 ural environment  can make attributing observed
 effects to specific  stressors difficult.
   The test data are used to quantify the relation-
 ship between the amount of the stressor and the
 magnitude of the response, and to evaluate the
 cause-effect relationship. Ideally, the stressor-
 response evaluation quantifies  the relationship be-
 tween the stressor and the assessment end  point.
 When the assessment end point can be measured,
 this analysis is straightforward. When it cannot be
 measured, the relationship between the stressor and
 measurement end point is established first, then ad-
ditional extrapolations, analyses, and assumptions
are used to predict or infer changes in the assess-
ment end point.

-------
                           Framework for ecological risk assessment at EPA
                                          1669
 between species, between responses, from labora-
 tory to field, and from field to field. Differences in
 responses among taxa depend on many factors, in-
 cluding physiology, metabolism, resource utiliza-
 tion, and life history. The relationship between
 responses also depends on many factors/including
 the mechanism of action and internal distribution
 of the stressor within the organism. When extrap- -
 olating between different laboratory  and field set-
 tings, important considerations include differences
 in the physical environment and organism behav-
 ior that will alter exposure, interactions with other
 stressors, and interactions with other ecological
 components.
     In addition to these extrapolations, an evalua-
 tion of indirect effects (e.g., effects on food, hab-
 itat, or competing species), effects at other levels of
 organization, effects at other temporal and spatial
 scales, and recovery potential may  be necessary.
 Whether these analyses are required in a particular
 risk assessment will depend on the assessment end
  points identified during problem formulation. The
  need for these types of analyses may also be iden-
  tified during risk characterization after an initial
  evaluation of risk.
     Another important aspect of the characteriza-
  tion of ecological effects is to evaluate the strength
  of the causal association between the stressor and
  the measurement and assessment end points. This
•  information supports and complements the stressor-
  response assessment and is of particular importance
  when the stressor-response relationship is based on
  field observations. An evaluation of causal evidence
  augments the risk assessment and contributes to the
  weight of evidence analysis supporting the judg-
  ment that a causal relationship exists between the
  stressor and response. Many of the concepts applied
   in human epidemiology can be useful for evaluat-
   ing observational  field studies [10]. An  example
   of ecological causality analysis  was provided by
   Woodman and Cowling [11], who evaluated the
   causal association between air pollutants and injury
   to forests.
      The results of the characterization of ecological
   effects are summarized in a stressor-response pro-
   file that describes the  stressor-response relation-
   ship, any extrapolations and additional analyses,
   and evidence of causality (e.g., field effects data).
   For practical reasons, the results of. the stressor-
    response analysis are often summarized as one ref-
   erence point, for instance, a 48-h LC50. Although
    useful, such values provide no information about
    rh^  <:lnne or shane of the stressor-resnonse curve.
points on the curve are identified, the difference in
magnitude of effect at different exposure levels can
be reflected in risk characterization.
   In our example, the characterization of ecolog-
ical effects may include stressor-response curves
from the laboratory or field that relate the concen-
trations of nutrients to changes in growth and re-
production of the plant species of interest. Studies
of nutrient loads to other, similar estuaries and the
associated response of vegetation may also be sum-
marized. The uncertainties and assumptions (e.g.,
that response in the field is the same as that in the
laboratory) would also be summarized.
 Risk characterization
    Risk characterization is the final phase of eco-
 logical risk assessment. The profiles of exposure and
 ecological effects serve as input to risk character-
 ization whenever risks are estimated and described.
    In the first step of risk characterization, risks
 are estimated by integrating the exposure and ef-
 fects data to yield an  expression of the likelihood
 of adverse effects occurring as a result of exposure
 to a certain stressor. Depending on the type of
 data, the risk may be expressed in a qualitative or
 quantitative fashion.  The integration may be per-
 formed by comparing single exposure and effect
 values, by comparing distributions within the ex-
 posure and effect profiles, or through the use of
 simulation models. The nature of the data and the
 requirements of the risk assessment will largely de-
 termine which method or combination of methods
 will be used. Another important activity associated
 with this step is the discussion of uncertainties en-
 countered during problem  formulation, analysis
  phase, and risk characterization. Uncertainties arise
  due to data and knowledge gaps, and the assessor
  must often use assumptions to bridge these gaps.
  Although such assumptions are necessary, their use
  often leads to uncertainty in tKe final assessment,
  which has to be acknowledged through a qualita-
  tive or quantitative uncertainty analysis.
      After the risks and uncertainties have been es-
  timated, the assessor summarizes the results and
  discusses the overall confidence in the risk assess-
  ment. This is achieved by objectively considering
  the sufficiency of the data, evidence of the cause-
  and-effect relationship, and any ancillary data in
  a weight-of-evidence evaluation. The objective is to
  describe the risk in terms of the assessment end
   point identified in the problem formulation phase.
   Without this crucial connection, the results of the
   risk assessment mav not be readily apparent to the

-------
 1670
S.B. NORTON ET AL.
 portant, the assessor provides an interpretation of
 the ecological significance of the identified risks.
 Ecological significance may be described in terms
 of the spatial and temporal extent of the effects; the
 nature and magnitude of the effects; and, when
 possible, an estimation of the recovery potential
 once the stressor is removed. Depending on the ob-
 jectives of the assessment, the risks may also  be
 placed in a broader ecological context by discuss-
 ing the implications of the effects to other compo-
 nents of the ecosystem.
    The risk characterization phase of our nutrient
 loading example would begin by integrating the nu-
 trient levels measured or modeled in the estuary
 with the stressor-response curves. These could be
 integrated simply by comparing different  points
 along the stressor-response curves with the concen-
 trations anticipated in the field. Alternatively, the
 stressor-response and exposure information could
 be integrated by using a model that would simulate
 the changes in the plant community resulting from
 the nutrient load. The  uncertainties of the assess-
 ment, including  those from the problem and anal-
 ysis phases, would be  described, and the overall
 confidence in  the assessment  would be discussed.
 The nature, magnitude, and spatial and temporal
 extent of effects would also be discussed. Finally,
 the effects on the plant community might be placed
 in a broader ecological context by discussing the
 implications for the wildlife and fish that depend
 on the plants.
   Risk characterization forms the basis for a dis-
 cussion of the results with the risk manager. The
 risk assessment is used by the  risk manager, along
 with economic,,  legal, and social concerns  in the
 risk management process, to evaluate management
 options. A consideration of the basic principles of
ecological risk assessment will contribute to a final
 product that is both credible  and germane to the
 needs of the risk manager.
   The "Framework Report" is the first step in a
 long-range effort to improve ecological risk assess-
ments within the EPA. In the short term, the basic
principles provided in the report are intended to
 foster a consistent approach for terminology in and
conduct of risk assessments.  The framework, as
part of a longer term guidelines development pro-
cess, will serve as a basis for identifying topics in
 future guidelines activities.

      FUTURE DEVELOPMENT OF EPA'S
 ECOLOGICAL RISK ASSESSMENT GUIDELINES
   The publication of  the "Framework Report"
          continuing the process into the second and third
          phases of guidelines development.
             The second phase of the strategy focuses on ac-
          quiring information on a series of guidance-issue
          areas that have been identified during several work-
          shops as essential to the development of a guideline
          [12,13]. The second phase of the program" will pro-
          duce a series of resource reports, comprised of one
          or more scientific white papers for each of the guid-
          ance-issue areas. In addition, a suite of problem-
          oriented  case studies will be developed to illustrate
          the application of the guidelines.
            The third phase of the program involves the in-
          tegration of the white papers, the problem-oriented
          examples, and the "Framework Report" into the
          first EPA-wide guidelines for ecological risk assess-
          ment. The goal of this plan is to use the framework
          as the platform for developing Agency ecological
          risk assessment guidelines by integrating new sci-
          entific information from each of the issue areas and
          case studies.

          Guidance-issue areas
            It became apparent during the process of devel-
         oping the "Framework Report" that there were a
         number of important scientific issues  for which
         additional  information and research  would be
         needed before ecological risk assessment guidelines
         could be developed. The following are the types
         of issues  and  needs that emerged from workshop
         discussions:

            Scale —the  issues related to spatial,  temporal,
              and biological scale
            Stressors —the need to define exposure for non-
              chemical stressors and multiple stressors
            End points — the importance of identifying the
              ecosystems and selecting the end points po-
              tentially at risk
            Ecological effects—the need for information on
              estimating direct, indirect, and cumulative ef-
              fects across biological scales
            Recovery—the importance of measuring the po-
              tential for ecosystem recovery
            Variability—the incorporation of estimates of
              natural variability into assessments
            Ecological significance—providing information
              on the ecological significance of change
            Uncertainty—treating uncertainty explicitly in
              risk assessments
           Causality—information on how  to determine
              causality in heterogeneously  stressed envi-
              ronments.

-------
                          Framework for ecological risk assessment at EPA
                                          1671
lustrate the risk assessment process as applied to
various types of problems. Second was the need for
risk management guidance that both identifies the
context within which the risk assessment resides
and discusses the interfaces between risk assessment
and risk management.

Problem-oriented case studies
   The strategy proposed for developing problem-
oriented case studies has two components: criteria
for selection and a format  for the analysis of the
case studies. The review and analysis of each case
study will follow the three-phase process outlined
in the framework. Specific attention will be given
to assessing the applicability of the process outlined
in the  framework  document and to the guidance
areas discussed  above. The case studies will also
provide valuable  information on analytical ap-
proaches, methodologies, and the scientific feasi-
bility of conducting risk assessments for particular
problem settings.
   The criteria used to select  the case studies in-
clude basic organizing principles (stressor type,
level of ecological organization, ecosystem type,
and spatial and temporal scale), the guidance-issue
areas, and regulatory needs.  For example, case
studies could be selected to illustrate toxic, non-
toxic chemical, and nonchemical stressors operat-
ing at widely different spatial and temporal scales
in different types of ecosystems.
   The value of the case studies to the ecological
risk assessment guideline process can be viewed
from several perspectives:
   Framework application—illustrate the applica-
     tion of the description of the risk assessment
     process to different types of problems
   Evaluation—evaluate the adequacy of the prin-
   •   ciples and concepts used in the framework
   Guidance areas-*-provide state-of-the-science in-
      formation on specific guidance issues being
      developed in parallel white papers
    Feasibility-determine the scientific feasibility
      (models, methods,- etc.) for conducting risk
      assessments  in a variety of problem settings
    Agency needs—provide specific examples of
      risk assessment to illustrate a spectrum of
      Agency regulatory needs.
 Preparation of ecological risk
 assessment guidelines
    The third phase of the proposed guideline pro-
 gram will involve  the preparation of ecological risk
            guidelines by the EPA. The issue-ori-
become the primary resources for the Agency's eco-
logical risk assessment guideline work groups. The
result will be a series of ecological risk assessment
guidelines that will include descriptions of principles
and concepts, provide generic guidance, include is-
sue- and problem-oriented resource volumes, and
contain specific applications of the risk assessment
process.
   The first document proposed for this phase is a
general ecological risk assessment guideline that
will provide specific guidance for the scientific is-
sues associated  with conducting ecological risk as-
sessments as outlined in the "Framework Report."
This approach is consistent with the scientific com-
munity's  recommendations to  develop an  initial
guideline that focuses on the scientific issues inte-
gral to the risk  assessment process.  The proposed
guideline, therefore, represents an expansion of the
principles and criteria developed in the framework.
The intention is that the ecological risk assessment
guideline provide detailed guidance and a range of
problem-oriented illustrations for each stage of the
risk assessment process. This approach assures that
the guideline will focus on the important scientific
issues and provide guidance on the use of the guide-
line for specific problems.

                  SUMMARY
   Increased awareness of ecological issues has
emphasized the need for improved ecological risk
assessment methodology. The  ecological risk  as-
sessment process will evolve as new ideas and re-
search advancements improve our basic knowledge
about how ecological components interact and how
stressors alter such interactions.
   The EPA has taken the first steps to develop
agencywide guidance for conducting ecological risk
 assessments, continuing a long-standing-EPA pro-
 gram to make the risk assessment process more sys-
 tematic. Inevitably, it will be a multiyear effort,
 just as it has been in the case of human health risk
 assessment, where development of those guidelines
 has been a product of several years of review and
 discussion involving scientists and  policy makers.
    Publication of the "Framework Report" now
 provides a structure on which future guidelines can
 be built. The EPA, however, recognizes that eco-
 logical risk assessment is a rapidly developing sci-
 ence driven by a desire to expand our capabilities
 beyond assessing single chemical effects on the re-
 sponses of individual  species. The framework has,
 therefore, been designed to accommodate a variety
 of stressors causins a diversity of ecological effects.

-------
  1672
                                           S.B. NORTON ET AL.
  ing regional- and global-scale problems increases.
  In its present form, the framework, through its dis-
  cussion of basic principles, operational definitions,
  and identification of key issues, will foster consis-
  tency  within the EPA and assist risk assessors in
  avoiding errors of omission or in pursuing risk as-
  sessment questions that cannot be applied in a reg-
  ulatory context.
     Over the next few years, more substantive guid-
  ance will be issued. Although the exact format for
  this guidance has not yet been decided, it will be an
  expansion of the principles discussed in the frame-
  work with emphasis being given to the specific el-
  ements of the risk assessment process, as described
  by the framework.  Just as in the case of the human
  health risk guidelines, these guidelines will not be
  rigid and will encourage the use of professional
 judgment within the framework  of a logical and
 scientifically sound structure. Development of risk
 assessment guidelines has historically provided an
 opportunity for discussion and debate among sci-
 entists, policy makers, and the public as to what
 society values and what level of protection is suf-
 ficient. This will certainly be true in the case of the
 ecological risk assessment guidelines, as the Agency
 grapples with translating its broad regulatory man-
 dates into concrete risk assessment policies and
 objectives.

.Acknowledgement—Suzanne Marcy cochaired the work
 group that developed early drafts of the "Framework Re-
 port." Other members of the work group included Mi-
 chael Brody, David Mauriello, Anne Sergeant, and Molly
 Whitworth. Many peer reviewers, both inside and outside
 the EPA, made valuable suggestions for revising  the
 "Framework Report." Especially noteworthy were  the
 participants in the May 1991 peer review workshop, which
 was chaired by James Fava, with discussions led by Law-
 rence Barnthouse, James Falco, Mark Harwell, and Ken-
 neth Reckhow.
10
11
12
13
                 REFERENCES
1. U.S. Environmental Protection Agency. 1992  A
   framework for ecological risk assessment  EPA
   630/R-92-001. Risk Assessment Forum, Washington,

2. U.S. Environmental Protection Agency. 1990. Re-
   ducing risk: Setting priorities and strategies for en-
   vironmental protection. SAB-EC-90-021. Science
   Advisory Board, Washington, DC.
3.  U.S. Environmental Protection Agency. 1992. Peer
   review workshop report on a "Framework for ecolog-
   ical risk assessment." EPA 625/3-91-022. Risk Assess-
   ment Forum, Washington, DC.
t.  Suter, G.W. II. 1990. Endpoints for regional ecolog-
   ical risk assessments. Environ. Manage. 14:9-23
>.  Suter, G.W. II.  1989. Ecological endpoints. In W
   Warren-Hicks, B.R. Parkhurst and S.S. Baker,  Jr.,
   eds., Ecological Assessments  of Hazardous Waste
   Sites: A Field and Laboratory Reference Document
   EPA 600/3-89-013. U.S. Environmental Protection
  Agency, Corvallis, OR, pp. 2-1-2-28.
i. Kelly, J.R. and M.A. Harwell. 1990. Indicators of
  ecosystem recovery. Environ. Manage. 14:527-546
. U.S. Department of Interior. 1987. Injury to fish and
  wildlife species. Type B Technical Information Doc-
  ument. CERCLA 301 Project. Washington, DC.
. U.S. Environmental Protection Agency. 1990. Eco-
  logical indicators. EPA 600/3-90-060. Environmental
  Monitoring and Assessment Program, Washington,

.  National Research Council. 1983. Risk assessment in
  the federal government: Managing the process. Na-
  tional Academy Press, Washington, DC.
.  Hill, A.B. 1965. The environment and disease: Asso-
  ciation or causation? Proc. R. Soc. Med. 58:295-300.
.  Woodman, J.N. and E.B. Cowling. 1987. Airborne
  chemicals and forest health. Environ. Sci. Techno/
  21:120-126.
  U.S. Environmental Protection Agency. 1992. Report
  on the ecological risk assessment guidelines strategic
 planning workshop.  EPA 630/R-92-002. Risk Assess-
 ment Forum, Washington, DC.
 U.S. Environmental Protection Agency. 1991. Sum-
 mary report on issues in ecological risk assessment.
 EPA 625/3-91-018.  Risk Assessment Forum, Wash-
 ington, DC.

-------
                      Susan v**i azouez-Tutt
     Susan Velasquez-Tutt received her Bachelor of Sciehoe degree



in Life sciences from the Massachusetts Institute of Technology in



W84.  She currently worfcs  for  the U.S.  EPA in the Environmental



Criteria and Assessment office in Cincinnati, Ohio and is pursuing



her doctoral degree in toxicology at the University of Cincinnati.



Her thesis research is an investigation into the activation of the



H-ras oncogene in liver tumors induced in male mice by chlorination



by-products  in drinKing water.   other areas of general  interest



include mechanisms of carcinogenesis and related  risK  assessment




issues,

-------
Combination of Cancer  Data  in Quantitative  Risk Assessments;

 Case Studies of Perchloroethylene  and  Bromodichloromethane
                       Sarah T. Vater1
                    Patricia M. McGinnis1
                    William  S.  Stiteler1
                   Susan  F. Velazquez-Tutt2
                       Linda A. Knauf2
                      Rita S.  Schoeny2


               1 Syracuse Research Corporation
                     2159 Gilbert Avenue
                   Cincinnati,  Ohio  45206


        Environmental Criteria and Assessment  Office
            U.S. Environmental Protection Agency
             26 W. Martin Luther King Jr. Drive
                   Cincinnati,  Ohio  45268
                     To  be  presented  at:
"Conference on the Risk Assessment Paradigm After Ten Years:
     Policy and Practice Then, Now, and in the Future"
                      April 5-8,  1993



              Wright-Patterson Air Force Base

                        Dayton,  Ohio

-------
                            ABSTRACT

     There  are often . several  data sets  that  may be  used  in
developing  a  quantitative risk estimate  for a carcinogen.   The
decision, however,  is usually made to base  this  estimate  on the
dose-response   data  for  tumor   incidences   from  a   single
sex/strain/species  of animal.   When appropriate,  the use of more
data  should result in a  higher level of confidence in  the risk
estimate.     The   decision  to  use  more   than  one  data  set
 (representing,  for  example,  different  animal  sexes,  strains,
 species  or  tumor sites)  can be  made  following  biological and
 statistical analyses of  the  compatibility  of  these  data  sets.
 Biological  analysis involves consideration of factors  such as the
 relevance of the animal models, study design  and  execution,  dose
 selection and route of  administration,  the mechanism of  action  of
 the agent,  its pharmacokinetics,  any species- and/or sex-specific
 effects, and  tumor site  specificity.   If the biological analysis
 does  not   prohibit   combining    data   sets,  the'  statistical
 compatibility  is then  investigated.  The generalized likelihood
 ratio method  is  proposed for determining  the  compatibility  of
 different  data sets with respect to a'common  dose-response model,
 such as the  linearized multistage model.    The  biological  and
 statistical  factors influencing the decision  to combine data sets
 are  described, followed by case-studies of perchloroethylene and
 bromodichloromethane.

-------
                            INTRODUCTION

      The estimation of the carcinogenic hazard posed to humans by

 a chemical involves a great deal of scientific judgement.
                            *
 Uncertainty is inherent to cancer  risk assessments,  particularly

 those developed from animal data, because assumptions must be made

 in areas for which data are  scarce.   These include,  for example,

 appropriate transformations  for extrapolation of dose  from test

 animals to humans  and for high-to-low doses,  and  the  assumption

 that  the  same  biological  process(es)   leading  to  cancer  in

 laboratory animals are operative in humans as well.   Statistical

 uncertainty is also inherent in the use of sample  data  to derive

 inferences about a large population.

      Oftentimes,  more than one statistically significant positive

 tumorigenic response is observed for  carcinogens,  either in both

 sexes in the same  bioassay,  or in a  different strain or species

 from a separate  study.   The  U.S.  EPA, however,  most  frequently

 bases carcinogen risk assessments  on results of a bioassay from a

 single sex/strain/species of animal (Stiteler  and Vater,  1989).

 The 1986  Guidelines for Cancer Risk Assessment (U.S. EPA,  1986a)

 suggest that the data set used in estimate quantitative cancer risk

.should be determined by the relevance of the animal  model to human

 health risk, the  quality of the data and the apparent sensitivity

 of the chosen animal  model.   In many cases,  however,  it  is  not

 known  whether  one   animal   model   is   more  appropriate   for

 extrapolation  to humans  than another.  When the differences  in

-------
animal sensitivities  are  small (as measured by  the quantitative
estimates derived from the data), or when the estimates incorporate
large uncertainties,  it could be  argued that a  more reasonable
approach  would be to derive a risk  estimate  using more  of the
available data.  A combination of data may result in a higher  level
of  associated  confidence  or it might result in a risk assessment
with  improved  statistical properties.
      While  the  use  of more of the  available  information  is  a
worthwhile  objective,  it  is complicated by  both biological  and
statistical issues.   The  available  carcinogenicity  data  for  a
 specific chemical rarely originate  from replicate studies;  rather,
 they are derived from  studies using  different sexes,  strains,
 and/or  species  of  animals in which  the responses depend  on  a
 variety  of biological  factors.   Differences  in  study design and
 execution may further complicate the issue of combining data from

 different sources.
      This work  was  initiated  to develop an approach in which more
 of the available data on carcinogenicity might be used to  derive a
 quantitative  risk estimate, as well as to determine when such  an
 approach would  be biologically  and statistically appropriate.
                    T» r-nMBTNTNG riPPTWOCmilCTTY INFORMATION
       There are  several  methods that could be  applied to utilize
  more of the available carcinogenicity information:  the choice of
  a risk  estimate'derived from a  single  data  set,  with additional

                                   3

-------
risk estimates from other data sets being used as corroboration for



the chosen value; the use of some average value (e.g.,  a median or



geometric mean) of risk estimates derived from different data sets;



or  the  combination   of  individual  data  sets,  prior  to  the



calculation of a risk estimate. - The term  "data set" is defined



here as the tumor incidence data for a  single anatomical site or



combination of sites within a single sex/strain/species of animal



(e.g.,  lung  adenomas  and carcinomas in  female F344  rats).  The



definition  of an  adequate data  set includes the  criterion  of



statistically  significant increased  tumor  incidence  in  exposed



groups by comparison to controls or by a trend analysis.



     If it is  assumed that the data sets  represent samples from the



same population,  then  statistically, the preferred  method  is to



combine them prior  to  derivation of  the quantitative estimate of



cancer potency.   This  is especially true  when the  estimate  is



expressed as an upper 95%  confidence limit on the slope in the low-



dose region of the animal dose-response  curve,  based on quantal



data utilizing a linearized multistage procedure.  As is  the case



for any confidence limit, this estimate reflects both the natural



variability and  the size  of the data set.   In general,  when all



other factors  are held  constant, a smaller data set yields a larger



value of the upper confidence limit than a larger data set because



of the uncertainty  inherent in basing the estimate  on a  limited



amount  of  information.     Thus,  the  average of  two  or  more



quantitative estimates each based on a small data set necessarily



incorporates   this  sample   uncertainty.     Alternatively,   the

-------
combination  of data  sets  prior  to  determining a  quantitative
estimate would decrease  this uncertainty and result  in a better
defined upper confidence level.
     This approach requires that the responses being considered for
combination be carefully  evaluated with respect to their biological
similarities and differences to judge whether a  combination of the
data is appropriate and would improve the overall risk assessment.
Examination  of the biological  basis for combination  is used in
conjunction  with  a statistical test  to evaluate the compatibility
of  two  or  more data sets with the  same multistage model,  that  is,
whether the  data  sets can be assumed to represent samples from  the
same population.  The statistical analysis consists of a likelihood
ratio test based  on information obtained from the GLOBAL86 version
 (Howe  et  al.,  1986) of the linearized multistage procedure.
       BIOLOGICAT.  CRITERIA FOR  COMPTNTNG  CARCINOGENTCITY  DATA
      The criteria for determining the biologic compatibility of 2
 data sets are discussed here as issues pertaining to study quality
 or to  the  mechanism of action  of the carcinogen.   Study quality
 factors  include  elements determined by the design of  the study
 itself  which  may affect  the  biological  processes  involved in
 carcinogenesis:   These  include study design  factors (e.g., dosing
 regimen  and  vehicle,   duration of exposure)  and  study quality
 factors  (e.g.,  purity  of test compound,  number  of  animals  and
 adequate survival).  Mechanistic  questions are those  pertaining

-------
directly to the assessment of how the chemical induces cancer,  such



as   genotoxicity,   species,   strain   or  sex   differences   in



pharmacokinetics  which  may affect the production of  carcinogenic



metabolites, or organ-specific responses  to the chemical.



     Figures 1  and 2 present these considerations in the form  of



decision trees that can be used in the determination  of whether  to



combine  data.    Several  qualitative  indicators  of   differential



sensitivity to a carcinogen (e.g.,   differences in latency, degree



of  malignancy)  may  also  be important.   Benign 'tumors  with the



potential  to progress  to malignancy  are usually considered  as



equivalent to malignant  tumors for purposes of quantitation.  While



benign and  malignant tumors are  frequently  combined within  data



sets, the decision to combine them across data sets may depend  on



the  tumor  type  and the  assumed  pattern  of  progression.    A



significant  difference  in  tumor  latency  might  be considered



evidence that  one sex,  strain  or species is substantially more



sensitive than another to the effects of-a carcinogen.



     Several of these biological  factors are considered  in more



detail in' the"  case  studies that  follow the  discussion of  the



statistical methodology.  Other biological issues not discussed in



these case  studies may  be pertinent  in the evaluation  of  other



chemicals.

-------
                                 Study Quality
        Yes
        Route of exposure
      same for all data sets?
                                                                No
Yes  -«
        Dose factors affecting
     assessment of dosen-esponse
      in one or more data sets?"
 Assess magnitude of
dose factor differences
                                                      Evidence for same systemic     ^
                                                       effects by multiple routes?      I
                     No
                                                   [JJ°J

                                                   DNC-
                                      Other study quality
                                          issues?"*
                              Will combining compensate
                                 for study deficiencies?    J
                                                                       ryi
                                Go to "Mechanism
                                    of Action"
               Do Not Combine                                           .
               «.g.,Dose rate (continuous vs. non-continuous); equivalent duration of
               exposure; MTD reached/exceeded
               .umber/size rf*»»g^ps;

 Figure 1. Decision tree used to analyze issues of study quality in combining cardnogenicity
                        risk assessment. These issues must be jcons.dered in
                          regarding the mechanism of action of the carcinogen.

-------
                            Mechanism of Action
Yes
Species - (sex, strain) - specific
 mechanism in any data set?
                                              No
                                        Chemical is gene-toxic?^}—^-
Yes

 T
                                                         Evidence for other similar  A
                                                      ^mechanism across data setsTy
                                                               „
                  Evidence for target organ
                   toxicity/cel! proliferation
                 in one data set (not others)?^/
                                                        I
            Do PB-PK models allow>
             target site dosimetry?
                                       Evidence for saturation of relevant
                                      metabolic pathways in one data set?
                             Yes
                                                                  No
                                                                  T
                                                   Is there tumor site concordance
                                                         across data sets?
          f
 ( Are tumors induced
  at multiple sites within
 \^ sex'strain/species?^
                       I
                                                      I
            (larget Site determined by:
              - chemical-specific factors
              - different kinetic/metabolic
               pathways
              - other host/organ-specific
            Vjactprs	,
                                                    'Do PB-PK models
                                                    support similarities
                                                     across data sets?,
                  Overall assessment of mechanistic similarity
                                                                             Yes
      *e.g., peroxiisome proliferation
 Figure 2. Decision tree used to analyze mechanistic similarity in carcinogenicity
 data sets to be considered for combination.
                                   8

-------
                                                       DATA
                                       CARCINOGENICITY
     The
         statistical test  proposed here is based  on generalized
likelihood ratio theory and will be illustrated in the context of
the  linearized  multistage  procedure.    The use  of  the  maximum
likelihood method for estimating unknown population parameter
related  to the  generalized  likelihood ratio  test  for  testing
                              likelihood estimation  of parameter
                                                             s is
hypotheses
thus,   maximum
estimates for  a
                                                               to
                 single data set will  be  presented here prior
               testing the compatibility of two data sets with the
discussions
same linearized multistage model
                                        DOSE-PKSPOWSE MODELS
              T.TKELIHOnn KSTIMATI™
      in a cancer bioassay,  where tumor development is considered to
 be a dichotomous response,  the number of exposed animals developing
 cancer could be expressed by the binomial probability distribution:
                                                              (1)
              fix)  =
                                    for x = 0,1,2, ••• ,N
 where N  is the number  of animals tested,  X is the  number with'
 tumors,  and P  is  the  probability  of developing  cancer.   The
 variables N and X  are established from the bioassay data, while the
 parameter P is unknown.   Substituting the sample values of X and N
 from the bioassay into  Equation 1, the value of P that maximizes
 the  equation can be estimated.   This value  of P represents the
                                "  9

-------
highest probability of observing that particular value of X.   From
this equation, the probability of observing any  number of animals
with cancer can be computed, so that Equation  1  also  represents a
likelihood function for the parameter P.  For an  experiment having
k groups  (including a control group and 2 or 3 dose groups) then,
the overall likelihood function can be  expressed as:
                                                              (2)
     For a given bioassay, the data points  X. and N;  that are input
into Equation 2 represent the total number  of animals with tumors,
and the number tested in the ith group, respectively.  In a well-
reported  study,  however,  individual  animal data  are available,
indicating a positive'or negative  carcinogenic response for each
animal.  Given  the dichotomous nature of the response, and assuming
the animals respond independently,  the likelihood function is then
expressed as a product  of Bernoulli distributions,  which,  except
for the exclusion of the combinatoric term,  is equivalent  to the
binomial-based likelihood:
                      L =
                                                              (3)
                                10

-------
     The likelihood function expressed  in  Equation  3  is simply a
product of terms  -  one
                        for each animal in the  experiment.   Each
tumor-bearing
              animal contributes  a  quantity P,  to  the likelihood
                                                           is the
and each nonresponder  contributes  a term (1-P,-) ,  where p.
probability  (unknown)  of  developing cancer for the  ith group to
which  the
           animal  belongs.   Looking  at this  problem with
                                                              the
Bernoulli distributions, then, means
                                     we
                                        can  specify the likelihood
solution  to
function  for combined  data  sets  from  two  separate  studies by
combining  the  individual  likelihood  functions.    The numerical
             the value of the P, terms is equivalent for both the
Binomial and Bernoulli problems.
     The problem remains,  then,  of how the dose  of  the carcinogen
affects the  likelihood function. This may be expressed by assuming
the data follow  a particular dose-response  model.   This has the
effect of   establishing .a  relationship  between  the P,
    tricting them to  functions that  include the measure of dose.
      To  illustrate,  we can assume the dose-response model to be a
    -stage model with no background rate:
                                                            terms,
 res
 one
                        P(d)  = 1 - exp(-6d) ,
                                                               (4)
 where  6  is an unknown  model parameter to be  estimated from the
 data.  For simplicity,  we can also assume that the higher of the
 two  doses  is  one unit  (d2=l)  and that the lower dose  is one-half
 that amount (d^O.5).   Then,  the probabilities of responding for
 each of  the two  groups  are:
                                 11

-------
                 ! = l - exp(-6 dj.)  = 1 - exp(-0.5 9) ,
and
P2  = 1 - exp(-6
                                       - exp(-9) .
Solving these two equations  for the common value 9 gives:
                           = -2ln(lt- Pi) ,
and
        9 = -
                                   - P7
These are set equal and  solved  to  give:
                        P2 = 1  -  (1 - Pt)
                                                               (5)
                                                               (6)
                                                               (7)
                                                               (8)
                                                               (9)
     The assumption of  a  one-stage  model  results in a restriction



of  the  maximum  of  the  likelihood  function  to  those  values



satisfying the equation, P2 = 1 -  (1  -  P.,)2.  All possible one-stage



models  are represented by this  equation.   The  maximum of  the



likelihood function under this constraint  occurs when P1 = 0.14 and



P2 = 0.26 (in contrast  a  perfect  fit would yield P1 = P2 = 0.20).



The value  of the model parameter,  9,  corresponding to  (P1,P2)  =



(0.14, 0.26) is 0.30, the maximum likelihood  estimate of 8.
         A STATISTICAL TEST OF COMPATIBILITY OF DATA SETS
                               ' 12

-------
   .  if  two  or  more data  sets  are  judged  to be  biologically

suitable for derivation of a quantitative estimate, the next step

is to determine if they  are  compatible with a common dose-response

model,  (i.e.,   the  same linearized  multistage   model).    This

determination could be done by testing the null hypothesis:



     H0: The data sets are compatible with a common model,
                               #


against the alternative hypothesis,



      H,: The data sets are not compatible with a common model.



      To test the null hypothesis,  the method of maximum likelihood,

described above,  can be used to estimate the model parameters.  If

two  or more data  sets  are  combined,   the resulting  likelihood

 function would be equal to  the product of the likelihood functions

 for  the individual data sets:
                             LT -
                                                             (10)
      If each of the individual terms is maximized, then it follows

 that  the  product  (i.e.,  the  overall  likelihood  function)  is

 maximized.  The joint  likelihood  LT can be maximized under either

 of two assumptions:  it can be assumed that the two  data sets can

 be fit to  a common dose-response  model  (H0) ; or it can be assumed

 that the two  data sets fit different  dose-response models  (H,) .

                                 13

-------
      In  order to test HQ, a generalized  likelihood ratio test can

be conducted by comparing the ratio of the two maxima under the two
assumptions:
                        A =
  maxL(ff0)
maxL(#0 U
                                                             (ID
Under  HQ,  -2 In  A  has  an  asymptotic  chi-square  distribution

 (Lindgren,  1976;  Cox and Hinkley,  1974).  The degrees  of  freedom

for  the  test are determined  by  the constraints on the parameter

space,  and  in  this case  are equal  to 1.   The  test, then,  is

performed by comparing -21n A = 2[max In L(HQ U H.,) - max In .L(HQ) ]

with the tabulated chi-square at a chosen level  of significance.

Alternatively,  we   could  use   the   chi-square   distribution  to

determine the probability  (p-value) of  seeing a  larger  value  of

-2In A than we have  calculated.

     We  would reject  HQ,  i.e.,  reject that the data sets  are

compatible  with  a  common  model and  accept  H1,  if  there  is  a

significant  difference  (p-value  <  chosen  significance  level)

between the values of the joint likelihood functions.when different

models are used for  the  two data sets and when  a common model is

used.  In contrast,  we would  accept H0, i.e.,  accept that the  two

data sets  are compatible  with  a common model  and reject H1, if

there is  little difference (p-value > chosen significance level)

between the maxima of the  joint likelihood functions under these

two assumptions.
                                14

-------
                 CASE STUTW l!
                           Background
     Tetrachloroethylene  (Perchloroethylene,  PCE)  is  a  volatile
chlorinated solvent which has been widely  used  as  a dry cleaning
agent and in industrial metal degreasing  operations.  PCE has been
reported to produce  increased incidence  cancer  in  rats and mice.
The data sets from these studies appeared to present possibilities
for  combination  both  because  of   their  apparent  quantitative
similarities  (McKone and Bogen, 1992) and the  fact that,  in the .
mouse, the same tumor site was observed in  both  sexes by different
routes  of administration.   In addition,  numerous  investigations
into the pharmacokinetics and mechanisms of action of the compound
have been reported.   The availability of  this  mechanistic animal
data,  as well as  data  on the pharmacokinetic behavior of PCE  in
humans,  afforded the opportunity to  examine a number of biological
 issues in more detail  than  is possible for most  chemicals.   The
 analysis for  PCE provides  an  example  of how certain types  of
 information can be  applied  in  the  process of determining whether
 data  sets  with  positive   responses  may   be  appropriate  for
 combination.
                       Discussion of  Data  Sets
       in the NCI  (1977)  study,  groups of Osborne-Mendel rats  and
  B6C3F1 mice (5o' animals/sex/dose group) were  administered  gavage
  doses of PCE (> 99% pure)  in corn oil,  5 days/week for 78 weeks.

                                  15

-------
Vehicle-treated  and  untreated  control  groups  consisted  of  so



animals/sex/species.   During  the  final 26  weeks,  treatment was



administered  to  rats  in  a pattern  of 1  week without  treatment



followed by 4  weeks with  treatment.   Rats and mice were  observed



for.  an • additional  32  and  12  weeks,  respectively,   following



treatment.  Time-weighted average doses in  (mg/kg)/day for the rats



were  471  and  941  for males  and 474 and  949 for  females.   The



corresponding  doses for  the mice are shown  in Table  1.   For the



mice, metabolized doses were calculated (U.S.  EPA,  1985)  based  on



estimates  of  total   urinary  metabolites,   derived  using  the



relationship described by Buben and O'Flaherty (1985).  Because  it



is widely accepted that the biological effects of  PCE are dependent



upon  its  metabolism' to more  reactive species,  the carcinogenic



response is assumed to be related to the metabolized dose.



     In the NCI  (1977) study, no  treatment-related increases  in



neoplastic lesions  were  reported in rats, however  this  may have



been  affected by  decreased survival  in   the  high-dose   animals.



Median survival times were 66 and 44 weeks  in  high-dose females and



males,  respectively,  versus  >88 weeks  in  controls.    In  mice,



significantly increased incidences of hepatocellular  carcinoma were



observed in mice  of both sexes at both dose levels,' as  compared



with their respective controls.  The  tumor  incidences are  shown  in



Table 1.




     An inhalation  bioassay in  F344/n  rats  and  B6C3F1  mice was



conducted by NTP  (1986).   Groups  of 50 animals/sex were exposed  to



PCE (99.9% pure)  6 hours/day,  5 days/week for 103  weeks, at levels






                               16

-------
of 0, 200  and  400 ppm  (rats) or  0,  100 and 200 ppm  (mice).   In
rats, the incidence of mononuclear cell  leukemia was significantly
increased  in rats of both sexes  at  both  exposure  levels: 28/50,
37/50,  37/50  in  males;  18/50,  30/50,  29/50  in females,  for
controls, 16w and high concentrations, respectively. Renal tubular
cell  adenomas  and carcinomas (combined),  a rare tumor type, also
occurred  in male  rats  with  a statistically significant  positive
trend  (1/49,  3/49,  4/50).    In mice,  significantly  increased
incidences  of   hepatocellular   carcinoma  alone  or   combined
adenoma/carcinoma were observed in treated animals  of both sexes at
both exposure levels.  The tumor incidences in mice  are shown in
Table 2.   For the inhalation study,  estimates'of metabolized dose
were based on  levels of urinary metabolites measured over a 72,hour
 period following a single six-hour inhalation exposure (Schumann et

 al., 1980; U.S. EPA, 1986b).
      Additional  studies which have  been  judged inadequate by the
 U.S. EPA  or which were not  positive have been reviewed  elsewhere

 (U.S. EPA, 1985).
      Several  data  sets  (consisting  of  anatomical sites  with
 statistically'significantly increased tumor incidence or a positive
 trend  in  tumor incidence) exist  that could potentially  be  used in
 the derivation of a quantitative risk estimate: hepatocellular
                                  17

-------
(0
        C
        o
        (d

       -P
        w









H
W
CQ
EH












(0
Q
Q)
CQ
C
O
a
W
0)
- 1
0)
CQ
0
Q
0)
Q)
r-H
• A
-p
CD
O
£_|
O
rH
A
U
rH
(U

Q)
(d
£>
(d
O

tp
C
•rH
o
rH
rH
O
Q)
O
•O
•H
O
G
H

JH
O
g

EH
       0)
       ca
       3
       O
       a
•o
tJ 3 C
Cy , ,_j
®3 s
° o
•p ^
(d d w .
II a |
w ^
1.1
•H < ^
5 \
g
TS ~
0) >i
N id
•H -3- rO
rH 
m s
S *~"
•d
0) •
w w \
X!
m iQ)
CQ *
!>i
fic3
o
CQ

O
CM


O

O






o





o

CO
CM
fO


CO
CM
f-
^
r>




en
o
VO





vo
CO
in

in
CM


en
fO
rH
CN
in




CM
CO





CM
o
&
M
-o

(d
•O
0)
-P
• X!
f"* C^
C*- -H
en a)
H ^
1
~ 0)
II
M M *
r*^
o



1











tumor.
                                                                                                                            CO
                                                                                                                            H

-------
CM
CQ
w
II ^
11 ^
1
EH
rd
rH
11 ^
it i
ii c— i
rH
NO}
3 °
3 *
rd a
4-J II Eg
tn
~ -rH 1
jS .s
Q "^
rd §
S C

C n
||




rd
• s
o
c
•H
O
rd
CJ



^^ 11 A fCJ ^^*
0 C C <
rn *-* 11 " f
c 2 c^-'
S •" ,3 rH
g .p 2 id r
p, TL n g JT
03 5 3 .2 J
^ 1 * i^
81 H |W:
UJ li
O i_ 1
Q % 	
Q) 's
s M --

5 £ |
S u <
§ s
rH 5
.£ 2
•^ EH 	
II
pi ® 1
•rH 1
CD 11
10
0
JS II
II


11
11
II
II
II
11
II


1


U it?
ill
JQD^
3 *-'


Q) ^*i
N (d
Hro 'CJ

Q 03 tJt
(D g

d
_o
Exposure
ncentrat:
(ppm)
o
o




4J

* *
•v* '^
nl ®

cn •*
13
0


**

VD






^







0






O

	


o



r-


H
CO





r-
•>*
in
CM







CTl




CM

a\
m




o
c
r-



0
m

o





o
in
VO
CM







H




in

CTl
in




o
o
CM









o^
M

~r^
0) n
rH 0
rd •
SO

S

•*





VO
rH







O






O




O'



CM
^

r-
H





CM
fl
rH







cn
CO




CM

n




o
o
H




"*

CO
m






1
n







in
•
n
H




in

c
in




o
o
CM



K!
•£
•
"fTrrt1
\o o
co XI
H X
«. 0)
rt! JH
w w
o
* c^
W X
. 0)
N_ tn
?**i
CD id
tn 13
o
T3 r**
T3 tO
/1\
b
r— I
0
(fl X5
• CD rd
XJ -P -P
vo rd 0)
co g B
cri -H
H 4-> H
U) rd
- Q) g
ril -r^
ft -PC
W O rd
Q)
rH II
W -H
•d <0
& C§
g OTJ
II o
M -d-P
	 
a . -o -H
>_ II rd vo Q) S
ff T3 co JJ o1
•.f1** rd cn rd 
-------
carcinoma  in  male and/or female  mice from gavage  administration



(NCI,  1977);  hepatocellular  carcinomas  alone or  combined with



adenomas in male and/or female mice from inhalation  exposure (NTP,



1986) ; mononuclear cell leukemia in male and/or female rats exposed



by  inhalation (NTP,  1986); and renal tubular  cell adenomas and



carcinomas in male rats exposed by inhalation  (NTP,  1986).



     The decision tree shown  in Figures 1 and 2 may be used as a



guideline for the analysis of  biological issues  which are relevant



to the question of whether to combine data sets for PCE.  The first



question to be considered is that of the route  of exposure  of the



chemical.  While the NCI (1977)  and NTP  (1986) administered  PCE by



gavage and inhalation,, respectively, the systemic effects in mice



were similar by both routes (hepatocellular carcinoma). There was



no  indication  of route-specificity  of  metabolism  or  effects,



probably in  part because PCE  is  not highly reactive and  has a



relatively long half-life in the  body.   The NCI (1977)  study did



use a shorter exposure  period  (78 weeks,  vs. 2 years  in NTP,  1986) ,



which introduces uncertainty into a quantitative combination  of the



data  sets.    Since  mice and  rats in the  NCI  (1977)  study were



observed for  up  to  90 and 110  weeks,  respectively, the  study



durations  were  similar.    Estimates  of  lifetime  average  daily



metabolized  dose (over  the  duration  of  the  study) have been



calculated based on estimates of urinary metabolites of PCE  (U.S.



EPA, 1985 and 1986b). These estimates, shown in Tables 1 and 2,



adjust in part for the differences between the exposure durations



and  the  dosing, regimens  in the  NCI  (1977)  and  the NTP   (1986)





                                20

-------
studies.   In this  particular  case,  the  question  of whether  a
combination of data across  routes  really presents advantages for
the risk assessment must  also  be addressed.   Route-specific risk
estimates,  for  oral and  inhalation exposure, may  be preferred,
since human exposure by both routes occurs.
     Since numerous studies have focused on the biochemical actions
of  PCE  and  related  compounds,  a more detailed examination of the
mechanisms  contributing  to tumor development is possible than  is
the case for most chemicals.  Referring to the Mechanism  of Action
section of the decision tree (Figure 2), the initial question to  be
considered  is whether  any of the data  sets represent species-  (or
strain  or sex)  specific  responses.  It  has been  suggested that the
kidney  tumors observed in  the in  the  male rat in  the NTP  (1986)
study may have been due at least in part  to the accumulation of the
protein a-2M-globulin following administration of  PCE  (Green  et
 al.,  1990), a response which is specific  to the male rat (reviewed
 in U.S.  EPA,  1991).   Subchronic  inhalation studies with  PCE  at
 exposure concentrations of 1000 ppm showed evidence of the hyaline
 droplet  nephropathy  which results  from accumulation  of  a-2/i-
 globulin  (Green et  al.,  1990).    However, direct histological
 evidence for this  mechanism at the doses used in  the NTP  (1986)
 study  is lacking.  Other hypotheses for the  induction of the renal
 tumors involve  renal  hydrolysis of  glutathione  conjugates   to
 mutagenic  metabolites (Green,  1990).  and regenerative hyperplasia
 associated with recurrent cytotoxicity.   Support  for  the  latter
 hypothesis includes evidence  of nephrotoxicity in rats in both  the

                                 21

-------
 NCI (1977)  and NTP (1986)  studies.   Whatever the actual mechanism



 or combination of  mechanisms, the probability of species and organ



 specificity suggests that the renal tumors  should  not  be  used in



 combination  with   other  data  sets  for  the  quantitative  risk



 assessment.
   *,



      The biological significance  of the mononuclear ,cell leukemia



 (MCL)  in rats of  both  sexes  has  also been questioned.   MCL  is  a



 common tumor in aging rats, with a high and variable spontaneous



 incidence.   In the NTP  (1986) study, the incidence of these tumors



 in concurrent controls  (56% in males;  36% in females) was unusually



 high compared  to  historical chamber controls  at  the  performing



 laboratory   (47%   in  males;   29%  in  females),  which raises  the



 questions regarding the increased incidence observed in  the study.



 The relevance to humans of  this tumor type has  also been debated.



 Thus,  while the PCE-induced MCL incidence may be indicative  of  a



 carcinogenic  response  in  the rat,  the use of  these  tumors  in



 combination  with  others  in  the  rat  or  the  mouse may  not  be



 appropriate.



      The use  of liver  tumor incidence  in B6C3F1 mice  for human



 health risk assessment  has  been questioned because  of the  genetic



 susceptibility of  this strain of  mice  to liver tumor  induction.



.The high spontaneous incidence is  believed to be  due in part  to  a



 genetic predisposition associated with a locus termed hcs (Hanigan,



 1990) ;   in  male mice  this  predisposition may be  compounded by



 hormonal influences (Haseman et  al., 1985).  However,  when liver



 tumors are  increased in mice of both sexes, and tumors are  seen at






                                22

-------
other sites or in other species, as is the case for PCE, it is less
likely that the response can be attributed solely to a sex/strain-
specific mechanism.  Therefore, data  sets based on the mouse liver
would not  be  excluded from combination  at this point  based on a
strain or  sex-specific mechanism.
     The next mechanistic question to be considered  is that.of the
genotoxicity of PCE. The evidence for genotoxicity of PCE, briefly
summarized here, is  equivocal (reviewed in  detail  in  U.S. EPA,
1985).   Results from  mutagenicity  assays,  in  the  presence or
absence of metabolic activation systems,  have been mostly  negative.
Studies  reporting  positive results  were  conducted at  cytotoxic
concentrations  and  showed  weak  responses.     In   addition,
mutagenicity   assays  conducted with highly  purified  PCE  showed
 negative results.   However, metabolites formed via a  glutathione
 conjugation pathway  may  be potential mutagens and have not  been
 thoroughly investigated.
      Sufficient data are  available  to demonstrate  significant
 differences  between  mice  and rats  in the  pharmacokinetics  and
 metabolism of PCE.   Several studies  have  shown  that  oxidative
 metabolism   of  PCE  reaches   saturation   at  relatively   low
 concentrations  in the  rat  (>  100  ppm) ,  relative to  the mouse
  (reviewed in U.S.  EPA,  1985; Bolt, 1987).  Thus, at  higher exposure
  levels, mice generate  higher levels of PCE metabolites than do
  rats. ' Trichloroacetic acid, a major metabolite of PCE,  is a potent
  peroxisome-inducing  agent, and  it has  been  suggested that  hepatic
  peroxisome proliferation might be a  causative factor in the genesis

                                  23

-------
of the mouse liver tumors (Odum et al., 1988).   The inability of



rats to generate sufficient levels of TCA from PCE exposure  (due to



saturation of the oxidative pathway) would explain why liver tumors



are  not seen  in  rats,  although a  causal  relationship  between



peroxisome proliferation and tumor induction has not been clearly



established.



     Overall,  the  lack  of   evidence  of  genotoxicity  and  the



demonstrated differences in metabolism  across species supports the



suggestion that the  two types of tumors observed in  the rat and the



liver tumors in the mice arise by different mechanisms and should



not be combined for the calculation of a quantitative estimate.



     For purposes  of combination, the available data sets appear to



consist of hepatocellular tumors in mice of both sexes,  by either



the oral or inhalation route.   Data sets  from male and female mice



were  modeled separately  for  the  gavage and  inhalation  studies



(Tables  1  and 2),  using the  linearized multistage  procedure.



Results of the  likelihood ratio tests for the male and female mouse



combined data sets  .(from both the oral and inhalation routes) are



shown in Table 3.  The p-value for  the  test of the null hypothesis



of compatibility  with a common model  is <0.05 for each combination



of data sets,  indicating rejection of  that hypothesis.  'Figure 3



depicts the plots of the multistage models  fit by Global86 for the



hepatocellular carcinoma incidence  in male and female mice, and for



the combined data sets,  from the NCI (1977)  study.   Similar plots



for the hepatocellular adenoma and/or carcinoma in individual and



combined data sets from the NTP (1986) study are shown in.Figure 4.





                                24

-------
in both cases, the differing shapes of the dose-response curves for



the males and females are evident, suggesting that sex-specific



differences  in  tumorigenesis  are  of  sufficient  magnitude  to



preclude combining the data sets.   Thus, the appropriate basis for



a quantitative risk estimate for PCE remains a single data set.
                                 25

-------
                           TABLE  3

                 Likelihood Ratio Tests for
       Perchloroethylene - Induced Mouse Liver Tumors
Study
NCI, 1977
(Gavage)
NTP,1986
(Inhalation)
NTP, 1986
(Inhalation)
Data Sets
cf ca1
9 ca
cf + 9 cal
cf ca
9 ca
cf + 9 ca
cf ad/ca2
9 ad/ca
cf + 9 ad/ca

3
2
6
3
1
5
q.,* p value
.4E-1 	
.5E-1 	 	
. 6E-1 	
.5E-1 	
	 <0.01
. 1E-1 	
.OE-2 	
	 <0.01
Compatible?
' No
No
No
'carcinoma only
2combined adenoma" and/or carcinoma
                             26

-------
                     FIGURE 3
 1,00
                                                             (1)
                                                              (2)
            10      30     30
                           DOSE
O


5

O
a
in
u
a


o
u

-------
                                FIGURE 4
        Dose-Response Curves for NTP (1986)  for Mouse Liver
Adenomas/Carcinomas: Males (1); Females  (2);  and Comi>ined(3)
             CD
             z

             5
             z
             o
             a
             en
             ui
             D;

             Z
             o
             a
             L_
             (D
             Z

             a

             o
             a
             (fi
             u
             a

             z
             o
             o
             c
             a
             L.
              CD
              Z

              a

              o
              a
              w
              u
              a

              z
              o
              o
              a:
                       »TP iimmtini srmv: frmf msr IM» mxal
(D
(2)
(3)
                                      10   12   U   16   18  M   2?
                                     28

-------
               CASE STUDY 2; BROMODTHH-LOROMETHANE
                           Background
 Bromodichloromethane  (BDCM)  is a volatile trihalomethane formed
when chlorine interacts with organic constituents in water.  BDCM
has been  detected in many untreated waters as  well  as drinking
water systems treated by chlorination (reviewed in U.S.  EPA, 1992).
Several  studies  have  suggested  a possible  association between
cancer  incidence,  particularly  of  bladder,   colon   and  rectal
cancers, with water chlorination  (U.S. EPA, 1992 and 1993).   There
are no epidemiological studies of BDCM intake alone, and the intake
of chlorinated  water  involves  exposure "to a mixture of compounds,
 including the trihalomethanes.
   Several animal'studies have attempted to evaluate whether chronic
 oral  exposure to BDCM can induce tumors.   Carcinogenicity studies
 wherein BDCM was administered in the drinking water (Tumasonis et
 al.,   1987;  Voronin   et  al.,   1987)   or  diet-supplemented  in
 microcapsules  (Tobe et al., 1982)  did  not induce a statistically
 elevated tumor incidence in the treated Wistar  rats or CBAxC57Bl/6
 mice  than  in  the  respective  controls.    These studies  have
 limitations  such as  incomplete histological examination, .lack  of
 discussion regarding solubility of BDCM in water at high doses  and'
 volatilization.   These  studies have been reviewed  in U.S.  EPA
  (1992).   NTP  (1987)  administered BDCM  by gavage  in  corn  oil  to
 F344/N rats and B6C3F1 mice  of both sexes and reported tumors at
 multiple sites.   Several  researchers  have  suggested that  the
  difference in  findings between the NTP and other studies- may be due
                                 " 29

-------
to different strains of rats,  the administration of a bolus dose

vs. continuous dosing and/or  an  interaction of the chemical with

the corn oil  vehicle.   The latter possibility has received much

attention with chloroform (Jorgenson et al.,  1985;  Bull et al.,

1986).

  The choice of data showing a statistically significant positive

tumorigenic  response  that  can be used  for a  quantitative risk

estimate are limited to those  from the NTP  (1987) bioassay.
                     Discussion of data sets
                        *
  NTP (1987) administered 0, 50 or 100 mg/kg/day BDCM (99% pure) in

corn  oil  to F344/N rats  (50/sex/dose)  by gavage  for 102 weeks.

Groups of 50 B6C3F1 mice received 0,  25  or  50 mg/kg/day  (males) or

0, 75 or 150 mg/kg/day  (females) BDCM by gavage in corn oil  for 102

weeks.  Controls received the vehicle.   The study in male rats was

restarted due to decreased survival of the vehicle  controls after

10.5  months due to excessive room  temperature.   Several sites

showing a,statistically significant  increased tumor incidence in

treated groups relative to controls or a statistically significant

dose-related  increase  in incidence  (positive  linear trend)  were

reported:   large  intestine  adenomatous  polyp/adenocarcinoma and

kidney tubular cell adenoma/adenocarcinoma in both  sexes of rats,

hepatocellular adenomas/adenocarcinomas in female mice,  and kidney

tubular  cell  adenoma/adenocarcinomas  in male  mice.    The tumor

incidences for these sites are shown in Tables 4-8.  The historical

vehicle  control, incidence for large intestine  tumors  and renal


                                30

-------
tubular cell tumors in male F344/N rats is 0.2% (3/1390) and 0.6%



(8/1448) ,  respectively,  and  in female rats of this  strain is 0%



(0/1400) and  0.1% (2/1447),  respectively  (NTP,  1987).   In male



B6C3F1  mice,  the historical  vehicle control  incidence of renal



tubular cell  tumors  is  0.3%  (5/1490).   The  vehicle  historical



control incidence of hepatocellular adenomas/carcinomas in  female



B6C3F1  mice  is 7.8%  (116/1489)  (NTP,  1987).   The low  historical



control  incidence  for  the   tumor  types  seen at  statistically



significant increased incidence in this study indicate these tumors



are uncommon and biologically important,  thereby constituting  data



sets adequate for calculation of a risk estimate (U.S. EPA,  1986a).



  Since all  the adequate  data sets are  from  the same bioassay,



issues  related to study quality are  minimal.  The route of exposure



for  both sexes  of  both species  was  oral  gavage  at  similar  dose



rates  and dose volumes  for lifetime of  the  animals.   Appropriate



controls were  employed, complete histology was  performed, and,  with



the  exception of the female mice, survival was comparable amongst



treated and' control groups.   While  survival in all  groups of the



 female mice was decreased due to ovarian abscesses, these data are



 not   compromised  for  use  in  quantitative   risk  estimation  as



 appropriate statistical analysis may adjust for survival.   The MTD
                                 31

-------
                             TABLE 4

       Bromodichloromethane Large Intestine Tumor Incidence
                          in F344/N Rats1
Human
Equivalent
Dose2 Adenomatous '
(m'g/kg) /day Polyp
Male Female
0.0 0/50 0/46
5.8 - 0/50
6.8 3/49
10.9 - " 7/47
13.0 33/50
Adeno-
carcinoma
Male
0/50
-
11/49
- .
3'8/50
Female
0/46
0/50
-
6/47
. -
Combined polyp
and carcinoma
Male Female
0/50 ' 0/46
0/50
13/49
12/47
45/50
1Adapted from NTP,  1987
2Based on surface area adjustment'
                                32

-------
                             TABLE 5

   Bromodichloromethane Renal Tumor Incidence  in  F344/N  Rats1
Human
Equivalent
Dose2       Tubular Cell
(mg/kg)/d    adenoma
                  Tubular Cell
                  carcinoma
                              Combined
                              adenoma
                              carcinoma
          Male   Female
                  Male
                                    Female
                                     Male
                                                        Female
   0.0

   5.8

   6.8

   10.9

   13.0
0/50
1/49
3/50
0/50

1/50




6/50
 0/50     0/50       0/50     0/50

         0/50        -       V50


 0/49      -        V49

         9/50        ~       15/50

10/50      -       13/50
      1 Adapted from NTP,  1987
      2Based on surface area adjustment
                                 33

-------
                             TABLE 6
        Bromodichloromethane Large Intestine and/or Renal
                  Tumor Incidence in F344/N Rats
Human
Equivalent
Dose2
(mg/kg)/day
0.0
5.8
6.8
10.9
13.0
Males "
0/50
	
13/49
	
46/50
Females
0/46
1/50
	
24/48
. 	
1 Adapted  from NTP,  1987
2Based  on surface area adjustment
                                34

-------
                             TABLE 7

Bromodichloromethane Liver Tumor Incidence in B6C3F1 Female Mice1
Human
Equivalent
Dose2
Hepatocellular
  adenoma
    adenoma  &
carcinoma
0.0
4.2
8.1
1/50
13/48
23/50
2/50
5/48
10/50
Combined

carcinoma

  3/50

  18/48

  29/50
   'Adapted from NTP, 1987
   2Based on surface area adjustment
                                 35

-------
                             TABLE  8

Bromodichloromethane Renal Tumor Incidence in B6C3F1 Male Mice1
Human
Equivalent
Dose2
(mg/kg)/day
0.0
1.5
3.0
Tubular Cell
adenoma
1/46
2/49
6/50
Tubular Cell
carcinoma
0/46
0/49
4/50
Combined adenoma
and carcinoma
1/46
2/49
9/50
1Adapted from NTP, 1987
2Based on surface area adjustment
                                36

-------
was reached in rats at the high dose as evidenced by decreased body
weight  and liver and  kidney  lesions.    Similarly,  histological
findings in the kidney, liver, thyroid and testis of the low dose
male mice and decreased body weight and thyroid hyperplasia  in high
dose female mice suggest the MTD was  reached.
     Concerns  about the use of corn oil as a  vehicle  in studies of
trihalomethanes have been raised (Withey et al., 1983; Jorgenson et
al.,  1985).   As such,  the Science Advisory Board of the U.S. EPA
recommended that the female mouse hepatocellular carcinomas not be
used  in quantitative risk assessment  (SAB, 1992).
      There  is  a  paucity   of  data   on  the  metabolism   and
pharmacokinetics  of BDCM,  limiting  comparisons  of mechanism  of
action among  sexes or species of animals.   Much of the information
 for  BDCM  has  been  inferred  from data   on  a   more  studied
 trihalomethane, chloroform.   In vivo and in  vitro studies with the
 trihalomethanes demonstrate that there are  two primary  routes of
 metabolism,  oxidative and reductive  (reviewed  in U.S.  EPA, 1992).
 Thorton-Manning et  al.   (1993)  recently demonstrated that P-450
 enzymes are involved in the metabolism of BDCM.   The research of
 Mink et al.  (1986) suggested that BDCM may be more rapidly absorbed
 and metabolized and/or  more  extensively metabolized'in mice than
 rats after a  single dose.
      •BDCM is considered by the U.S. EPA to have genotoxic potential
  (U.S.  EPA,  1992).'  Conflicting results in vivo and in  vitro test
  systems have  been attributed  to  inadequacies or  variation  in
  experimental protocols, such as difficulty in achieving  sufficient

                                 37

-------
  exposure to volatile  BDCM and different metabolic  capability of



  various cell types (NTP,  1987; reviewed in U.S. EPA, 1992).  Based



  on the  premise  that BDCM  is  most likely  to act by  a genotoxic



  mechanism,  and that similar genetic processes  form  the basis for



  carcinogenicity,  all data sets, with the exception of female mouse



  liver tumors, thus  far qualify as  the  basis  for a quantitative



  estimate of risk.



       Renal  tubular hyperplasia was reported in both sexes of rats.



  In addition, renal cytomegaly was noted in male rats and male mice



  at the high dose levels (NTP,  1987).   These findings suggest that



  an epigenetic mechanism, such as regenerative  hyperplasia, may also



  play a role  in carcinogenic activity of BDCM at this site, although



  the data are. insufficient to establish such a role.   Reitz et al.



  (1982) suggested  this  mechanism  for chloroform  carcinogenesis.



  The NTP reported that  hyaline droplet formation in the kidney was



  not seen in  the male rats, therefore, involvement of a-2ju-globulin



  nephrbpathy in the  formation of  these  tumors appears unlikely.



  Hyperplasia was  not  reported in  the large intestine  of either sex



  or species  (NTP,  1987).             '



       It is difficult to assess the role differential toxicokinetics



  among  the   sexes  or   species   may  have  on  the  mechanism . of



  carcinogenicity   of  BDCM  for  these  data sets.   Differences  in



.  absorption  (e.g., residence  time in the gastrointestinal tract) or



  metabolic rates and/or  pathways may account for the observation of



  large  intestine  tumors  in rats  but not  in  mice,  and  in  the



  observation of renal tumors  in male mice and hepatocellular tumors





                                 38

-------
in female mice.   The interaction of the vehicle  (corn  oil)  with
BDCM may  also  be a  determinant  in site-specificity/sensitivity,
although  the Science  Advisory  Board  of  the U.S.  EPA did  not
consider a vehicle effect  likely  in the development of the renal or
large intestine tumors (SAB,  1952).' without definitive biological
data, there do not appear to be mechanistically different processes
occurring at the different sites or between the different sexes or
species    that    would    discourage   combining    across   the
sites/sexes/species.   Therefore, possibilities for calculation of
a quantitative risk  estimate are:  male rat large  intestine/kidney
tumors,   female  rat  large   intestine/kidney tumors  and  large
 intestine/kidney tumors  in both sexes  combined.
      The linearized multistage procedure was used to model  these
 data sets and  the likelihood ratio test applied to evaluate the
 compatibility of the combined data set of both sexes as described
 above.  Table 9 shows that the data sets for male and female rats
 are not  statistically compatible  with the same  multistage model.
 These  results  suggest  that  some  of  the  biological  factors
  influencing the dose-response relationship are not evident  from the
  available information. "  Reexamination of the large intestine and
  renal tumor incidence shows that large intestine tumors occur  at.
  the  low dose  in the male rats but not in the female rats.   Thus,
  the   large  intestine/kidney  tumor" incidence  in  the males   is
  comprised primarily  of  large intestine  tumors,  whereas,  in the
  females kidney and large intestine tumors contribute equally to the
  combined incidence.  This could implicate such  factors as

                                 39

-------
                             TABLE  9

           Bromodichloromethane  Likelihood Ratio Tests
Data Sets
tf Mouse renal
9 Rat renal
d1 Rat renal
9 Rat Gl/renal
cf Rat Gl/renal
cf + 9 Rat Gl/renal
cf + 9 Rat renal
*
q.
6.2E-2
9.5E-3
8.5E-3
l.OE-2
2.5E-2.
	
6.3E-3
p value . Compatible?
	 	
	 	
	 	
	 • 	
	 	
<0.001 No
0.067 Yes
cf + 9 Rat renal and
    cf Mouse renal
<0.. 001
No
                                40

-------
differential metabolic pathways (e.g.,  oxidative in the kidney and
reductive in  the intestine)  in. the  disparate  responses observed
across sexes, although  further research  is  needed to clarify the
underlying biological basis.
  The observation of renal tubular cell adenomas  and adenocarcinomas
in  both  sexes  of  rats and the  male mice may  have biological
relevance and  tends  to support  the  hypothesis  that  a similar
mechanism of  action may be operating at this site.  However, the
renal tumors  in mice occurred at lower dose  levels than  the tumors
in  rats,  suggesting the mice may be "more sensitive" than the  rats.
Moore et al.  (1993) showed that nephrotoxicity occurs to a greater
extent    in   mice   than  in   rats   when  BDCM   is   administered
subchronically in drinking water.  The  most conservative option for
quantitative estimate  is  the  use  of only  the  male mice  kidney
 tumors   (Table 9).   Biological  similarity of the tumor type  argues
 in  favor of  combining  data sets across both sexes of  rats  and
 possibly across both rats and mice.   The latter  options have the
 advantage  of increasing the  number of  dose  groups to  five  and
 seven,   respectively.   The  results,  as presented  in Table 9, show
 that the male and female rat kidney tumor data are compatible with
 the  same multistage  model.    When  these combined data are tested
 with the male mice kidney data, the likelihood ratio test indicates
 these data sets are not compatible with the same  multistage model.
   in summary,  the  risk assessor is  left with various  options  to
  estimate the carcinogenic  risk  to BDCM.   The most conservative
  option is  the  use  of the male mouse kidney data alone.   Use  of  the

                                 41

-------
        large intestine/kidney tumors for each sex of rat and kidney tumors



        for male and  female rats combined are two other  options which are



        both biologically and statistically  viable.  Use of each tumor site



        in each  sex of rat alone provide four  other options.   The slope



        factors  for these eight data sets  are  shown  in  Table  9.   All of



        these  quantitative estimates are  within an  order of magnitude.



        Biological  information  has  provided little to resolve which data



        set is most appropriate, leaving the decision to the judgement of



        the assessor.








                                    DISCUSSION



          Our confidence  in  quantitative cancer risk  assessments may be



        increased by  the use of as many of the available  data as possible.



        Case studies on perchloroethylene and bromodichloromethane are used



        to illustrate many of the biological and statistical  issues that



        must  be  considered  in combining  multiple  data  sets  used  to



        calculate a cancer risk estimate.  In the case of PCE, bioassays in



        mice and rats were available by the oral route and by  inhalation.



        Several  data  sets demonstrating   a  statistically  significantly



        increased tumor incidence were available:  hepatocellular tumors in



        male and female mice by both routes  of administration;  mononuclear



        cell  leukemia  in  both sexes of  rats  by  inhalation;  and kidney



        tumors in male rats by inhalation.  An investigation into  biologic



        issues, however, reveals that the mechanisms by which these tumors



        are  likely to arise  are sufficiently  different to preclude the



        combination of data sets, except  for  the liver tumors  in  male and






                                        42
_

-------
female mice.   Subsequent statistical analysis of  the  mouse data
sets for both the gavage and  inhalation studies indicated that the
data from  the two sexes could  not  be combined.    In the  case of
bromodichloromethane,  only  one  NTP  bioassay  was  available for
quantitative  analysis.   In this-experiment, BDCM was administered.
by gavage  to  rats and mice,  with a positive tumor response  being
observed  for the  large  intestine  and kidney of  male  and  female
rats, liver tumors in female mice,  and kidney tumors in male mice.
Study design  issues (i.e.,  use of corn oil gavage)  precluded use of
the  liver  tumors.   The  remaining  data  sets  seemed plausible
candidates for  combination,  and   statistical  compatibility was
subsequently analyzed.   While the  kidney tumor  incidences in male
and  female   rats were  found  to   be compatible  with  the  same
multistage  model,  the  resulting   potency  estimate was  less
 conservative than that  generated from the male mouse  kidney data
 alone.   Based  on these findings,  options for risk management  of
 BDCM  are presented.    The final  decision,  as  with  all  risk
 assessments, will  involve the use  of assumptions and scientific
 j udgement.
                                 43

-------
                            REFERENCES

H.M.  Bolt,  "Pharmacoki.netic Factors and Their  Implication  in  the
Induction of Mouse Liver Tumors by Halogenated Hydrocarbons," Arch.
Toxicol., Suppl.  10,  190-203  (1987).


J.R.  Buben, E.J. O'Flaherty, "Delineation of the Role of Metabolism
in  the  Hepatotoxicity of Trichloroethylene and Perchloroethylene:
.A Dose-Effect Study," Toxicology and  Applied  Pharmacology
78,  105-122  (1985).

B.J.  Bull,   J.M.  Brown,  E.A.  Meierhenry,  T.A.  Jorgenson,   M.
Robinson,  and J.A. Stober, "Enhancement of the Hepatotoxicity of
Chloroform in B6C3F1  Mice by Corn Oil: Implications for chloroform
Carcinogenesis,"   Environmental   Health  Perspectives  69,   49-58
 (1986).

Cox,  D.R. Cox and D.V. Hinkley, Theoretical Statistics. Chapman and
Hall, New York,  NY (1974).

T.  Green,   "Chloroethylenes:  A Mechanistic Approach to
Human  Risk   Evaluation,"    Annual  Review  of  Pharmacology  and
Toxicology  30, 73-89  (1990).

T.Green, J.  Odum, and J.K.  Foster,  "Perchloroethylene-Induced Rat
Kidney  Tumors: An Investigation of  the Mechanisms  Involved  and
Their Relevance  to Humans,"  Toxicology and  Applied Pharmacology
 103,  77-89  (1990).

M.H.   Hanigan,   M.C.   Winkler  and  N.R.  Drinkwater,   "Partial
hepatectomy is a  promoter of hepatocarcinogenesis in C57BL/6J male
mice but not  in  C3H/HeJ male mice," Carcinogenesis,  11, 589-594
 (1990).

J.K.  Haseman, J.E. Huff, G.N. Rao, J.E.  Arnold,  G.A.  Boorman and
 E.E.  McConnell,  "Neoplasms observed  in  untreated and  corn  oil
 gavage  control groups  of F344/N rats and  (C57BL/6N x C3H/HeN)Fl
 (B6C3F1)  mice,"  J.  Natl. Cancer Inst., 75, 975-984 (1985).

 R.B.  Howe,  K.S.  Crump, and  C. Van Landingham,  "GLOBAL86.    A
 computer program to extrapolate quantal animal toxicity data to low
 doses," K.S. Crump & Company, Ruston LA (1986).

 T.A.  Jorgenson, E.F.  Meierhenry,  C.J. Rushbrook, R.J.  Bull, and M.
 Robinson,  "Carcinogenicity of Chloroform in Drinking Water to Male
 Osborne-Mendel  Rats   and Female  B6C3F1  Mice," Fundamental  and
 Applied Toxicology 5, 760-769 (1985).

 B.W. Lindgren, Statistical Theory. 3rd  Ed..  MacMillan Publishing
 Co.," Inc.  New York,  NY  (1976).
                                 44

-------
T.E.  McKone  and   K.T.   Bogen,   "Uncertainties  in  Health-Risk
Assessment:  An Integrated Case Study Based on Tetrachloroethylene
in California Ground Water," Regulatory Toxicology and Pharmacology
15, 86-103 (1992).

P.L. Mink, J. Brown, and J. Rickabaugh, "Absorption, Distribution
and Excretion of 14C-trihalomethanes in Mice and Rats," Bulletin of
Environmental Contamination and Toxicology 37,  752-758  (1986).

T.C. Moore, A.B. DeAngelo,  and R.A\ Pegram, "Subchronic Toxicity of
Broraodichloromethane and Bromoform Administered to Mice and Rats in
Drinking  Water," The  Toxicologist 13(1),  359,  (Abstract 1405)
(1993) .

National  Cancer Institute  (NCI),  "Bioassay of  Tetrachloroethylene
for   Possible  Carcinogenicity",  U.S.   Department  of   Health,
Education, and  Welfare Publication No.  (NIH) 77-813  (1977).

National Toxicology Program, "Toxicology and Carcinogenesis Studies
of  Tribromomethane  (Bromoform)  in F344/N  Rats  and  B6C3F1  Mice
(Gavage Studies),"  (U.S.  Department of Health  and Human Services,
Research  Triangle  Park,  N.C.,  Technical Report Series No.  321,
1987) .

National  Toxicology  Program,  "Toxicology  and  Carcinogenesis  of
Tetrachloroethylene (Perchloroethylene in F344/N Rats  and B6C3F1
Mice  (Inhalation Studies)," NIH  Publication  No. 86-2567,  NTP  TR
311,  (1986).

J.  Odum., T.  Green, J.R.  Foster, and  P.M.   Hext,  "The  role  of
trichloroacetic  acid   and  peroxisome   proliferation   in   the
differences  in carcinogenicity of perchloroethylene in the mouse
and rat," Toxicology and Applied Pharmacology 92, 103-112 (1988).

R.H.  Reitz,  T.R. Fox,  and J.F. Quast,  "Mechanistic  Considerations
for Carcinogenic Risk Estimation," Chloroform Environmental Heath
Perspective 46, 163-168, (1982).

SAB.  (Science Advisory Board) , Drinking Water Committee, "Review of
Health Criteria Document for Trihalomethanes by the Drinking water-
Committee of the  Science Advisory  Board."   (U.S.  Environmental
Protection Agency Science Advisory Board, Washington,  D.C., 1992)

A.M.  Schumann,  T.F. Quast, P.G. Watanabe, "The  Pharmacokinetics of
 Perchloroethylene  in  Mice and  Rats  as  Related to Oncogenicity,
 Toxicology and Applied Pharmacology 55, 207-219  (1980).

 W. Stiteler, and  S.  Vater, "Improved Methods for  Combination of
 Quantitative Risk Estimates for Carcinogens",   Prepared  by Syracuse
 Research Corporation, Syracuse,  NY under Contract  No.  68-C8-0004
 for   the  Office   of  Health   and  Environmental   Assessment,
 Environmental  Criteria  and  Assessment  Office, Cincinnati,  OH.

                                 45

-------
(1989).

J.R.   Thorton-Manning,   P.    Gao,    and   P.O.   Lilly,   "Acute
Bromodichloromethane Toxicity  in  Rats Pretreated With Cytochrome
P450  Inducers  and  Inhibitors,"  The  Toxicologist  13(1),  361
(Abstract 1412) (1993).

M. Tobe, Y.  Suzuki,  K.  Aida, H. Yoshimoto,  K.  Takada, J. Kadoma, K.
Kobayashi, Y. Vehida,  K. Anda, Y. Nakaji,  Y. Suzuki, and M. Sato,
"Studies  on   the   Chronic  Oral  Toxicity  of  Tnbromomethane,
Dibromochloromethane   and   Bromodichloromethane,"   (Unpublished
Intraagency Report to the National Institute  of Hygienic Sciences,
Tokyo, Japan, Tokyo Medical and Dental University,  (1982).

C.F. Tumasonis, D.N. McMartin, and B. Bush,  "Lifetime Toxicity of
Chloroform  and  Bromodichloromethane  When  Administered  Over   a
Lifetime  In Rats,"  Ecotoxicology  and  Environmental Safety  9,  233-
240  (1985).

U.S. Environmental  Protection  Agency,  "Health Assessment  Document
for  Tetrachloroethylene   (Perchloroethylene)",   Prepared  by the
Office  of  Health  and  Environmental  Assessment,  Environmental
Criteria  and  Assessment Office,  Research  Triangle Park, NC.  EPA
600/8-82-005F (1985).

U.S.  Environmental  Protection Agency, "Guidelines  for  Carcinogen
Risk Assessment,"   Federal Register  51(185)  33992-34003  (1986a).

U.S.  Environmental  Protection Agency,  "Addendum  to  the  Health
Assessment  Document for Tetrachloroethylene  (Perchloroethylene)
 (Prepared  by the  Office  of  Health  Environmental  Assessment,
Washington,   D.C.,  EPA/600/8-82/005FA. Available  from:  National
Technical Information Service, Springfield,  VA;  PB-86-174489/AS),
 (1986b).

U.S. Environmental Protection Agency, "Alphas-Globulin: Association
With Chemically Induced Renal Toxicity and  Neoplasia in  the Male
Rat."  Prepared for the Risk Assessment Forum, U.S. Environmental
Protection  Agency,  Washington D.C. (1991).

U.S. Environmental  Protection Agency,  "Drinking Water  Criteria
 Document for Trihalomethanes.   Revised  External Review  Draft,"
 Office of Science and Technology, Washington,  D.C., (1992).

 U.S. Environmental  Protection Agency,  "Integrated Risk Information
 System,.Online,  Office of  Health and Environmental  Assessment,
 Environmental Criteria  and  Assessment Office,  Cincinnati,  Ohio,
 (1993).
 W.M. Voronin,  A.I. Donchenko, and A.A. Korolev, "Experimental Study
 of    the    Carcinogenicity    of    Dichlorobromomethane    and

                                 46

-------
Dibroraochloromethane Which Are Formed During the Water Chlorination
Process," Gigiena  i Sanitariia (Eng.  Trans.) 19-21 (1987).

W.M. Voronin and N.N.  Litvirov,  "Carcinogenicity of Chloroform in
the Mouse," Voprosy onkolog 33(8),  81-85 (1987).

J.R. Withey, B.T. Collins, P.G. Collins, "Effect of Vehicles on the
Pharmacokinetics and Uptake of Four Halogenated Hydrocarbons From
the  Gastrointestinal  Tract  of  the   Rat,"  Journal   of  Applied
Toxicology  3,  249-253  (1983).
                                 47
                                                 •fc OS. GOVERNMENT PWHTWO OFFICE: 1983-751-979

-------